Решение практической задачи 2

Задача: прогнозирование потребительских предпочтений на промышленные товары

Решение практической задачи выполняется в два этапа. На первом этапе необходимо использовать библиотеку OCEAN-AI для получения гипотез предсказаний (оценок персональных качеств личности человека). На втором этапе следует использовать метод _priority_calculation из библиотеки OCEAN-AI для решения представленной практической задачи. Примеры результатов работы и реализации представлены ниже.

Таким образом, библиотека OCEAN-AI предоставляет инструмент для анализа персональных качеств личности потребителей, что полезно для предсказания того, что их заинтересует. Это позволит компаниям более точно адаптировать свои товары и услуги к предпочтениям потребителей, делая их более уникальными и персонализированными.

71a2d9673a194f169ab636d1f96e5e2a

3ad3499d44ff4bcdbba16c8c3f0d03a9

`FI V2`

[2]:

# Импорт необходимых инструментов
import os
import pandas as pd

# Импорт модуля
from oceanai.modules.lab.build import Run

# Создание экземпляра класса
_b5 = Run()

# Настройка ядра
_b5.path_to_save_ = './models' # Директория для сохранения файла
_b5.chunk_size_ = 2000000      # Размер загрузки файла из сети за 1 шаг

corpus = 'fi'

# Формирование аудиомоделей
res_load_model_hc = _b5.load_audio_model_hc()
res_load_model_nn = _b5.load_audio_model_nn()

# Загрузка весов аудиомоделей
url = _b5.weights_for_big5_['audio'][corpus]['hc']['googledisk']
res_load_model_weights_hc = _b5.load_audio_model_weights_hc(url = url, force_reload = False)

url = _b5.weights_for_big5_['audio'][corpus]['nn']['googledisk']
res_load_model_weights_nn = _b5.load_audio_model_weights_nn(url = url, force_reload = False)

# Формирование видеомоделей
res_load_model_hc = _b5.load_video_model_hc(lang='en')
res_load_model_deep_fe = _b5.load_video_model_deep_fe()
res_load_model_nn = _b5.load_video_model_nn()

# Загрузка весов видеомоделей
url = _b5.weights_for_big5_['video'][corpus]['hc']['googledisk']
res_load_model_weights_hc = _b5.load_video_model_weights_hc(url = url, force_reload = False)

url = _b5.weights_for_big5_['video'][corpus]['fe']['googledisk']
res_load_model_weights_deep_fe = _b5.load_video_model_weights_deep_fe(url = url, force_reload = False)

url = _b5.weights_for_big5_['video'][corpus]['nn']['googledisk']
res_load_model_weights_nn = _b5.load_video_model_weights_nn(url = url, force_reload = False)

# Загрузка словаря с экспертными признаками (текстовая модальность)
res_load_text_features = _b5.load_text_features()

# Формирование текстовых моделей
res_setup_translation_model = _b5.setup_translation_model() # только для русского языка
res_setup_translation_model = _b5.setup_bert_encoder(force_reload = False)
res_load_text_model_hc_fi = _b5.load_text_model_hc(corpus=corpus)
res_load_text_model_nn_fi = _b5.load_text_model_nn(corpus=corpus)

# Загрузка весов текстовых моделей
url = _b5.weights_for_big5_['text'][corpus]['hc']['googledisk']
res_load_text_model_weights_hc_fi = _b5.load_text_model_weights_hc(url = url, force_reload = False)

url = _b5.weights_for_big5_['text'][corpus]['nn']['googledisk']
res_load_text_model_weights_nn_fi = _b5.load_text_model_weights_nn(url = url, force_reload = False)

# Формирование модели для мультимодального объединения информации
res_load_avt_model_b5 = _b5.load_avt_model_b5()

# Загрузка весов модели для мультимодального объединения информации
url = _b5.weights_for_big5_['avt'][corpus]['b5']['googledisk']
res_load_avt_model_weights_b5 = _b5.load_avt_model_weights_b5(url = url, force_reload = False)

PATH_TO_DIR = './video_FI/'
PATH_SAVE_VIDEO = './video_FI/test/'

_b5.path_to_save_ = PATH_SAVE_VIDEO

# Загрузка 10 тестовых аудиовидеозаписей из корпуса First Impression V2
# URL: https://chalearnlap.cvc.uab.cat/dataset/24/description/
domain = 'https://download.sberdisk.ru/download/file/'
tets_name_files = [
    '429713680?token=FqHdMLSSh7zYSZt&filename=_plk5k7PBEg.003.mp4',
    '429713681?token=Hz9b4lQkrLfic33&filename=be0DQawtVkE.002.mp4',
    '429713683?token=EgUXS9Xs8xHm5gz&filename=2d6btbaNdfo.000.mp4',
    '429713684?token=1U26753kmPYdIgt&filename=300gK3CnzW0.003.mp4',
    '429713685?token=LyigAWLTzDNwKJO&filename=300gK3CnzW0.001.mp4',
    '429713686?token=EpfRbCKHyuc4HPu&filename=cLaZxEf1nE4.004.mp4',
    '429713687?token=FNTkwqBr4jOS95l&filename=g24JGYuT74A.004.mp4',
    '429713688?token=qDT95nz7hfm2Nki&filename=JZNMxa3OKHY.000.mp4',
    '429713689?token=noLguEGXDpbcKhg&filename=nvlqJbHk_Lc.003.mp4',
    '429713679?token=9L7RQ0hgdJlcek6&filename=4vdJGgZpj4k.003.mp4'
]

for curr_files in tets_name_files:
    _b5.download_file_from_url(url = domain + curr_files, out = True)

# Получение прогнозов
_b5.path_to_dataset_ = PATH_TO_DIR # Директория набора данных
_b5.ext_ = ['.mp4'] # Расширения искомых файлов

# Полный путь к файлу с верными предсказаниями для подсчета точности
url_accuracy = _b5.true_traits_[corpus]['googledisk']

_b5.get_avt_predictions(url_accuracy = url_accuracy, lang = 'en')

[2024-10-10 18:10:50] Извлечение признаков (экспертных и нейросетевых) из текста …

[2024-10-10 18:10:50] Получение прогнозов и вычисление точности (мультимодальное объединение) …

10 из 10 (100.0%) … GitHub:nbsphinx-math:OCEANAI\docs\source\user_guide:nbsphinx-math:notebooks\video_FI:nbsphinx-math:test_plk5k7PBEg.003.mp4 …

	Path	Openness	Conscientiousness	Extraversion	Agreeableness	Non-Neuroticism
Person ID
1	2d6btbaNdfo.000.mp4	0.618917	0.660694	0.477656	0.654437	0.601256
2	300gK3CnzW0.001.mp4	0.461732	0.413451	0.415706	0.498301	0.431224
3	300gK3CnzW0.003.mp4	0.468002	0.448618	0.371742	0.509602	0.453739
4	4vdJGgZpj4k.003.mp4	0.585348	0.616446	0.49443	0.605614	0.587017
5	be0DQawtVkE.002.mp4	0.680991	0.56602	0.553915	0.646545	0.64246
6	cLaZxEf1nE4.004.mp4	0.66342	0.551018	0.557912	0.585238	0.587174
7	g24JGYuT74A.004.mp4	0.590237	0.399273	0.409554	0.531861	0.507134
8	JZNMxa3OKHY.000.mp4	0.60577	0.523617	0.531137	0.594406	0.57984
9	nvlqJbHk_Lc.003.mp4	0.511002	0.464702	0.390882	0.443663	0.438811
10	_plk5k7PBEg.003.mp4	0.647606	0.610466	0.524718	0.61428	0.606428

[2024-10-10 18:10:50] Точность по отдельным персональным качествам личности человека …

	Openness	Conscientiousness	Extraversion	Agreeableness	Non-Neuroticism	Mean
Metrics
MAE	0.0735	0.0631	0.0914	0.0706	0.0691	0.0735
Accuracy	0.9265	0.9369	0.9086	0.9294	0.9309	0.9265

[2024-10-10 18:10:50] Средняя средних абсолютных ошибок: 0.0735, средняя точность: 0.9265 …

Лог файлы успешно сохранены …

— Время выполнения: 35.449 сек. —

[2]:

True

Для прогнозирования потребительских предпочтений в промышленных товарах необходимо знать коэффициенты корреляции, определяющие взаимосвязь между персональными качествами личности человека и предпочтениями в товарах или услугах.

В качестве примера предлагается использование коэффициентов корреляции между персональными качествами человека и характеристиками автомобилей, представленными в статье:

O’Connor P. J. et al. What Drives Consumer Automobile Choice? Investigating Personality Trait Predictors of Vehicle Preference Factors // Personality and Individual Differences. – 2022. – Vol. 184. – pp. 111220.

Пользователь может установить свои коэффициенты корреляции.

Прогнозирование потребительских предпочтений на характеристики атомобиля

[3]:

# Загрузка датафрейма с коэффициентами корреляции
url = 'https://download.sberdisk.ru/download/file/478675818?token=EjfLMqOeK8cfnOu&filename=auto_characteristics.csv'
df_correlation_coefficients = pd.read_csv(url)
df_correlation_coefficients = pd.DataFrame(
    df_correlation_coefficients.drop(['Style and performance', 'Safety and practicality'], axis = 1)
)
df_correlation_coefficients.index.name = 'ID'
df_correlation_coefficients.index += 1
df_correlation_coefficients.index = df_correlation_coefficients.index.map(str)

df_correlation_coefficients

[3]:

	Trait	Performance	Classic car features	Luxury additions	Fashion and attention	Recreation	Technology	Family friendly	Safe and reliable	Practical and easy to use	Economical/low cost	Basic features
ID
1	Openness	0.020000	-0.033333	-0.030000	-0.050000	0.033333	0.013333	-0.030000	0.136667	0.106667	0.093333	0.006667
2	Conscientiousness	0.013333	-0.193333	-0.063333	-0.096667	-0.096667	0.086667	-0.063333	0.280000	0.180000	0.130000	0.143333
3	Extraversion	0.133333	0.060000	0.106667	0.123333	0.126667	0.120000	0.090000	0.136667	0.043333	0.073333	0.050000
4	Agreeableness	-0.036667	-0.193333	-0.133333	-0.133333	-0.090000	0.046667	-0.016667	0.240000	0.160000	0.120000	0.083333
5	Non-Neuroticism	0.016667	-0.006667	-0.010000	-0.006667	-0.033333	0.046667	-0.023333	0.093333	0.046667	0.046667	-0.040000

[4]:

_b5._priority_calculation(
    correlation_coefficients = df_correlation_coefficients,
    col_name_ocean = 'Trait',
    threshold = 0.55,
    number_priority = 3,
    number_importance_traits = 3,
    out = False
)

_b5._save_logs(df = _b5.df_files_priority_, name = 'auto_characteristics_priorities_fi_en', out = True)

# Опционно
df = _b5.df_files_priority_.rename(columns = {'Openness':'OPE', 'Conscientiousness':'CON', 'Extraversion': 'EXT', 'Agreeableness': 'AGR', 'Non-Neuroticism': 'NNEU'})
columns_to_round = ['OPE', 'CON', 'EXT', 'AGR', 'NNEU']
df[columns_to_round] = df[columns_to_round].apply(lambda x: [round(i, 3) for i in x])
df

[4]:

	Path	OPE	CON	EXT	AGR	NNEU	Priority 1	Priority 2	Priority 3	Trait importance 1	Trait importance 2	Trait importance 3
Person ID
1	2d6btbaNdfo.000.mp4	0.619	0.661	0.478	0.654	0.601	Safe and reliable	Practical and easy to use	Economical/low cost	Conscientiousness	Agreeableness	Openness
2	300gK3CnzW0.001.mp4	0.462	0.413	0.416	0.498	0.431	Classic car features	Fashion and attention	Luxury additions	Agreeableness	Conscientiousness	Openness
3	300gK3CnzW0.003.mp4	0.468	0.449	0.372	0.510	0.454	Classic car features	Fashion and attention	Luxury additions	Agreeableness	Conscientiousness	Openness
4	4vdJGgZpj4k.003.mp4	0.585	0.616	0.494	0.606	0.587	Safe and reliable	Practical and easy to use	Economical/low cost	Conscientiousness	Agreeableness	Openness
5	be0DQawtVkE.002.mp4	0.681	0.566	0.554	0.647	0.642	Safe and reliable	Practical and easy to use	Economical/low cost	Agreeableness	Conscientiousness	Openness
6	cLaZxEf1nE4.004.mp4	0.663	0.551	0.558	0.585	0.587	Safe and reliable	Practical and easy to use	Economical/low cost	Conscientiousness	Agreeableness	Openness
7	g24JGYuT74A.004.mp4	0.590	0.399	0.410	0.532	0.507	Classic car features	Recreation	Luxury additions	Agreeableness	Conscientiousness	Non-Neuroticism
8	JZNMxa3OKHY.000.mp4	0.606	0.524	0.531	0.594	0.580	Practical and easy to use	Safe and reliable	Economical/low cost	Agreeableness	Openness	Non-Neuroticism
9	nvlqJbHk_Lc.003.mp4	0.511	0.465	0.391	0.444	0.439	Classic car features	Fashion and attention	Luxury additions	Agreeableness	Conscientiousness	Openness
10	_plk5k7PBEg.003.mp4	0.648	0.610	0.525	0.614	0.606	Safe and reliable	Practical and easy to use	Economical/low cost	Conscientiousness	Agreeableness	Openness

Прогнозирование потребительских предпочтений на характеристики мобильного устройства

В качестве примера предлагается использование коэффициентов корреляции между персональными качествами человека и характеристиками мобильного устройства, представленными в статье:

Peltonen E., Sharmila P., Asare K. O., Visuri A., Lagerspetz E., Ferreira D. (2020). When phones get personal: Predicting Big Five personality traits from application usage // Pervasive and Mobile Computing. – 2020. – Vol. 69. – 101269.

[5]:

# Загрузка датафрейма с коэффициентами корреляции
url = 'https://download.sberdisk.ru/download/file/478676690?token=7KcAxPqMpWiYQnx&filename=divice_characteristics.csv'
df_divice_characteristics = pd.read_csv(url)

df_divice_characteristics.index.name = 'ID'
df_divice_characteristics.index += 1
df_divice_characteristics.index = df_divice_characteristics.index.map(str)

df_divice_characteristics

[5]:

	Trait	Communication	Game Action	Game Board	Game Casino	Game Educational	Game Simulation	Game Trivia	Entertainment	Finance	Health and Fitness	Media and Video	Music and Audio	News and Magazines	Personalisation	Travel and Local	Weather
ID
1	Openness	0.118	0.056	0.079	0.342	0.027	0.104	0.026	0.000	0.006	0.002	0.000	0.000	0.001	0.004	0.002	0.004
2	Conscientiousness	0.119	0.043	0.107	0.448	0.039	0.012	0.119	0.000	0.005	0.001	0.000	0.002	0.002	0.001	0.001	0.003
3	Extraversion	0.246	0.182	0.211	0.311	0.102	0.165	0.223	0.001	0.003	0.000	0.001	0.001	0.001	0.004	0.009	0.003
4	Agreeableness	0.218	0.104	0.164	0.284	0.165	0.122	0.162	0.000	0.003	0.001	0.000	0.002	0.002	0.001	0.004	0.003
5	Non-Neuroticism	0.046	0.047	0.125	0.515	0.272	0.179	0.214	0.002	0.030	0.001	0.000	0.005	0.003	0.008	0.004	0.007

[6]:

_b5._priority_calculation(
    correlation_coefficients = df_divice_characteristics,
    col_name_ocean = 'Trait',
    threshold = 0.55,
    number_priority = 3,
    number_importance_traits = 3,
    out = True
)

_b5._save_logs(df = _b5.df_files_priority_, name = 'divice_characteristics_priorities_fi_en', out = True)

# Опционно
df = _b5.df_files_priority_.rename(columns = {'Openness':'OPE', 'Conscientiousness':'CON', 'Extraversion': 'EXT', 'Agreeableness': 'AGR', 'Non-Neuroticism': 'NNEU'})
columns_to_round = ['OPE', 'CON', 'EXT', 'AGR', 'NNEU']
df[columns_to_round] = df[columns_to_round].apply(lambda x: [round(i, 3) for i in x])
df

[6]:

	Path	OPE	CON	EXT	AGR	NNEU	Priority 1	Priority 2	Priority 3	Trait importance 1	Trait importance 2	Trait importance 3
Person ID
1	2d6btbaNdfo.000.mp4	0.619	0.661	0.478	0.654	0.601	Game Casino	Game Educational	Game Trivia	Non-Neuroticism	Conscientiousness	Agreeableness
2	300gK3CnzW0.001.mp4	0.462	0.413	0.416	0.498	0.431	Media and Video	Entertainment	Health and Fitness	Conscientiousness	Agreeableness	Extraversion
3	300gK3CnzW0.003.mp4	0.468	0.449	0.372	0.510	0.454	Media and Video	Entertainment	Health and Fitness	Conscientiousness	Agreeableness	Extraversion
4	4vdJGgZpj4k.003.mp4	0.585	0.616	0.494	0.606	0.587	Game Casino	Game Educational	Game Trivia	Non-Neuroticism	Conscientiousness	Agreeableness
5	be0DQawtVkE.002.mp4	0.681	0.566	0.554	0.647	0.642	Game Casino	Communication	Game Trivia	Non-Neuroticism	Extraversion	Agreeableness
6	cLaZxEf1nE4.004.mp4	0.663	0.551	0.558	0.585	0.587	Game Casino	Communication	Game Trivia	Non-Neuroticism	Extraversion	Agreeableness
7	g24JGYuT74A.004.mp4	0.590	0.399	0.410	0.532	0.507	Health and Fitness	Media and Video	Entertainment	Openness	Conscientiousness	Agreeableness
8	JZNMxa3OKHY.000.mp4	0.606	0.524	0.531	0.594	0.580	Game Casino	Game Educational	Game Simulation	Non-Neuroticism	Agreeableness	Openness
9	nvlqJbHk_Lc.003.mp4	0.511	0.465	0.391	0.444	0.439	Media and Video	Entertainment	Health and Fitness	Agreeableness	Conscientiousness	Extraversion
10	_plk5k7PBEg.003.mp4	0.648	0.610	0.525	0.614	0.606	Game Casino	Game Educational	Game Trivia	Non-Neuroticism	Agreeableness	Conscientiousness

Прогнозирование потребительских предпочтений по стилю одежды

В качестве примера предлагается использование коэффициентов корреляции между персональными качествами человека и стилем одежды, представленными в статье:

Stolovy T. Styling the self: clothing practices, personality traits, and body image among Israeli women // Frontiers in psychology. - 2022. - vol. 12. - 719318.

[7]:

# Загрузка датафрейма с коэффициентами корреляции
url = 'https://download.sberdisk.ru/download/file/493644097?token=KGtSGMxjZtWXmBz&filename=df_%D1%81lothing_style_correlation.csv'
df_clothing_styles = pd.read_csv(url)

df_clothing_styles.index.name = 'ID'
df_clothing_styles.index += 1
df_clothing_styles.index = df_clothing_styles.index.map(str)

df_clothing_styles

[7]:

	Trait	Comfort	Camouflage	Assurance	Fashion	Individuality
ID
1	Openness	0.01	-0.24	0.31	0.07	0.31
2	Conscientiousness	-0.03	-0.24	0.17	0.09	0.15
3	Extraversion	-0.01	-0.19	0.30	0.13	0.14
4	Agreeableness	0.16	-0.16	0.15	-0.09	-0.05
5	Non-Neuroticism	0.03	-0.16	0.01	0.00	0.06

[8]:

_b5._priority_calculation(
    correlation_coefficients = df_clothing_styles,
    col_name_ocean = 'Trait',
    threshold = 0.55,
    number_priority = 3,
    number_importance_traits = 3,
    out = True
)

_b5._save_logs(df = _b5.df_files_priority_, name = 'clothing_styles_priorities_fi_en', out = True)

# Опционно
df = _b5.df_files_priority_.rename(columns = {'Openness':'OPE', 'Conscientiousness':'CON', 'Extraversion': 'EXT', 'Agreeableness': 'AGR', 'Non-Neuroticism': 'NNEU'})
columns_to_round = ['OPE', 'CON', 'EXT', 'AGR', 'NNEU']
df[columns_to_round] = df[columns_to_round].apply(lambda x: [round(i, 3) for i in x])
df

[8]:

	Path	OPE	CON	EXT	AGR	NNEU	Priority 1	Priority 2	Priority 3	Trait importance 1	Trait importance 2	Trait importance 3
Person ID
1	2d6btbaNdfo.000.mp4	0.619	0.661	0.478	0.654	0.601	Assurance	Individuality	Comfort	Openness	Conscientiousness	Agreeableness
2	300gK3CnzW0.001.mp4	0.462	0.413	0.416	0.498	0.431	Camouflage	Fashion	Comfort	Conscientiousness	Openness	Non-Neuroticism
3	300gK3CnzW0.003.mp4	0.468	0.449	0.372	0.510	0.454	Camouflage	Fashion	Comfort	Conscientiousness	Openness	Non-Neuroticism
4	4vdJGgZpj4k.003.mp4	0.585	0.616	0.494	0.606	0.587	Assurance	Individuality	Comfort	Openness	Conscientiousness	Agreeableness
5	be0DQawtVkE.002.mp4	0.681	0.566	0.554	0.647	0.642	Assurance	Individuality	Fashion	Openness	Extraversion	Conscientiousness
6	cLaZxEf1nE4.004.mp4	0.663	0.551	0.558	0.585	0.587	Assurance	Individuality	Fashion	Openness	Extraversion	Conscientiousness
7	g24JGYuT74A.004.mp4	0.590	0.399	0.410	0.532	0.507	Camouflage	Individuality	Fashion	Agreeableness	Openness	Non-Neuroticism
8	JZNMxa3OKHY.000.mp4	0.606	0.524	0.531	0.594	0.580	Comfort	Individuality	Assurance	Openness	Agreeableness	Non-Neuroticism
9	nvlqJbHk_Lc.003.mp4	0.511	0.465	0.391	0.444	0.439	Camouflage	Comfort	Fashion	Conscientiousness	Openness	Non-Neuroticism
10	_plk5k7PBEg.003.mp4	0.648	0.610	0.525	0.614	0.606	Assurance	Individuality	Comfort	Openness	Conscientiousness	Agreeableness

`MuPTA` (ru)

[9]:

import os
import pandas as pd

# Импорт модуля
from oceanai.modules.lab.build import Run

# Создание экземпляра класса
_b5 = Run()

corpus = 'mupta'
lang = 'ru'

# Настройка ядра
_b5.path_to_save_ = './models' # Директория для сохранения файла
_b5.chunk_size_ = 2000000      # Размер загрузки файла из сети за 1 шаг

# Формирование аудиомоделей
res_load_model_hc = _b5.load_audio_model_hc()
res_load_model_nn = _b5.load_audio_model_nn()

# Загрузка весов аудиомоделей
url = _b5.weights_for_big5_['audio'][corpus]['hc']['googledisk']
res_load_model_weights_hc = _b5.load_audio_model_weights_hc(url = url, force_reload = False)

url = _b5.weights_for_big5_['audio'][corpus]['nn']['googledisk']
res_load_model_weights_nn = _b5.load_audio_model_weights_nn(url = url, force_reload = False)

# Формирование видеомоделей
res_load_model_hc = _b5.load_video_model_hc(lang=lang)
res_load_model_deep_fe = _b5.load_video_model_deep_fe()
res_load_model_nn = _b5.load_video_model_nn()

# Загрузка весов видеомоделей
url = _b5.weights_for_big5_['video'][corpus]['hc']['googledisk']
res_load_model_weights_hc = _b5.load_video_model_weights_hc(url = url, force_reload = False)

url = _b5.weights_for_big5_['video'][corpus]['fe']['googledisk']
res_load_model_weights_deep_fe = _b5.load_video_model_weights_deep_fe(url = url, force_reload = False)

url = _b5.weights_for_big5_['video'][corpus]['nn']['googledisk']
res_load_model_weights_nn = _b5.load_video_model_weights_nn(url = url, force_reload = False)

# Загрузка словаря с экспертными признаками (текстовая модальность)
res_load_text_features = _b5.load_text_features()

# Формирование текстовых моделей
res_setup_translation_model = _b5.setup_translation_model() # только для русского языка
res_setup_translation_model = _b5.setup_bert_encoder(force_reload = False)
res_load_text_model_hc_fi = _b5.load_text_model_hc(corpus=corpus)
res_load_text_model_nn_fi = _b5.load_text_model_nn(corpus=corpus)

# Загрузка весов текстовых моделей
url = _b5.weights_for_big5_['text'][corpus]['hc']['googledisk']
res_load_text_model_weights_hc_fi = _b5.load_text_model_weights_hc(url = url, force_reload = False)

url = _b5.weights_for_big5_['text'][corpus]['nn']['googledisk']
res_load_text_model_weights_nn_fi = _b5.load_text_model_weights_nn(url = url, force_reload = False)

# Формирование модели для мультимодального объединения информации
res_load_avt_model_b5 = _b5.load_avt_model_b5()

# Загрузка весов модели для мультимодального объединения информации
url = _b5.weights_for_big5_['avt'][corpus]['b5']['googledisk']
res_load_avt_model_weights_b5 = _b5.load_avt_model_weights_b5(url = url, force_reload = False)

PATH_TO_DIR = './video_MuPTA/'
PATH_SAVE_VIDEO = './video_MuPTA/test/'

_b5.path_to_save_ = PATH_SAVE_VIDEO

# Загрузка 10 тестовых аудиовидеозаписей из корпуса MuPTA
# URL: https://hci.nw.ru/en/pages/mupta-corpus
domain = 'https://download.sberdisk.ru/download/file/'
tets_name_files = [
    '477995979?token=2cvyk7CS0mHx2MJ&filename=speaker_06_center_83.mov',
    '477995980?token=jGPtBPS69uzFU6Y&filename=speaker_01_center_83.mov',
    '477995967?token=zCaRbNB6ht5wMPq&filename=speaker_11_center_83.mov',
    '477995966?token=B1rbinDYRQKrI3T&filename=speaker_15_center_83.mov',
    '477995978?token=dEpVDtZg1EQiEQ9&filename=speaker_07_center_83.mov',
    '477995961?token=o1hVjw8G45q9L9Z&filename=speaker_19_center_83.mov',
    '477995964?token=5K220Aqf673VHPq&filename=speaker_23_center_83.mov',
    '477995965?token=v1LVD2KT1cU7Lpb&filename=speaker_24_center_83.mov',
    '477995962?token=tmaSGyyWLA6XCy9&filename=speaker_27_center_83.mov',
    '477995963?token=bTpo96qNDPcwGqb&filename=speaker_10_center_83.mov',
]

for curr_files in tets_name_files:
    _b5.download_file_from_url(url = domain + curr_files, out = True)

# Получение прогнозов
_b5.path_to_dataset_ = PATH_TO_DIR # Директория набора данных
_b5.ext_ = ['.mov'] # Расширения искомых файлов

# Полный путь к файлу с верными предсказаниями для подсчета точности
url_accuracy = _b5.true_traits_['mupta']['sberdisk']

_b5.get_avt_predictions(url_accuracy = url_accuracy, lang = lang)

[2024-10-10 18:20:29] Извлечение признаков (экспертных и нейросетевых) из текста …

[2024-10-10 18:20:30] Получение прогнозов и вычисление точности (мультимодальное объединение) …

10 из 10 (100.0%) … GitHub:nbsphinx-math:OCEANAI\docs\source\user_guide:nbsphinx-math:notebooks\video_MuPTA:nbsphinx-math:test\speaker_27_center_83.mov …

	Path	Openness	Conscientiousness	Extraversion	Agreeableness	Non-Neuroticism
Person ID
1	speaker_01_center_83.mov	0.765745	0.696637	0.656309	0.75986	0.494141
2	speaker_06_center_83.mov	0.686514	0.659488	0.611838	0.749739	0.420672
3	speaker_07_center_83.mov	0.671993	0.661216	0.571759	0.704542	0.381026
4	speaker_10_center_83.mov	0.69828	0.59893	0.571893	0.674907	0.35082
5	speaker_11_center_83.mov	0.718329	0.598986	0.573518	0.73201	0.379845
6	speaker_15_center_83.mov	0.670932	0.671055	0.602337	0.708656	0.399527
7	speaker_19_center_83.mov	0.767261	0.658167	0.653367	0.801366	0.463443
8	speaker_23_center_83.mov	0.699837	0.684907	0.616671	0.806437	0.447853
9	speaker_24_center_83.mov	0.710566	0.66299	0.610562	0.711242	0.413696
10	speaker_27_center_83.mov	0.759404	0.712562	0.658357	0.830507	0.507612

[2024-10-10 18:20:30] Точность по отдельным персональным качествам личности человека …

	Openness	Conscientiousness	Extraversion	Agreeableness	Non-Neuroticism	Mean
Metrics
MAE	0.0706	0.0788	0.1328	0.1071	0.1002	0.0979
Accuracy	0.9294	0.9212	0.8672	0.8929	0.8998	0.9021

[2024-10-10 18:20:30] Средняя средних абсолютных ошибок: 0.0979, средняя точность: 0.9021 …

Лог файлы успешно сохранены …

— Время выполнения: 324.067 сек. —

[9]:

True

Для прогнозирования потребительских предпочтений в промышленных товарах необходимо знать коэффициенты корреляции, определяющие взаимосвязь между персональными качествами личности человека и предпочтениями в товарах или услугах.

В качестве примера предлагается использование коэффициентов корреляции между персональными качествами человека и характеристиками автомобилей, представленными в статье:

O’Connor P. J. et al. What Drives Consumer Automobile Choice? Investigating Personality Trait Predictors of Vehicle Preference Factors // Personality and Individual Differences. – 2022. – Vol. 184. – pp. 111220.

Пользователь может установить свои коэффициенты корреляции.

Прогнозирование потребительских предпочтений на характеристики атомобиля

[10]:

# Загрузка датафрейма с коэффициентами корреляции
url = 'https://download.sberdisk.ru/download/file/478675818?token=EjfLMqOeK8cfnOu&filename=auto_characteristics.csv'
df_correlation_coefficients = pd.read_csv(url)
df_correlation_coefficients = pd.DataFrame(
    df_correlation_coefficients.drop(['Style and performance', 'Safety and practicality'], axis = 1)
)
df_correlation_coefficients.index.name = 'ID'
df_correlation_coefficients.index += 1
df_correlation_coefficients.index = df_correlation_coefficients.index.map(str)

df_correlation_coefficients

[10]:

	Trait	Performance	Classic car features	Luxury additions	Fashion and attention	Recreation	Technology	Family friendly	Safe and reliable	Practical and easy to use	Economical/low cost	Basic features
ID
1	Openness	0.020000	-0.033333	-0.030000	-0.050000	0.033333	0.013333	-0.030000	0.136667	0.106667	0.093333	0.006667
2	Conscientiousness	0.013333	-0.193333	-0.063333	-0.096667	-0.096667	0.086667	-0.063333	0.280000	0.180000	0.130000	0.143333
3	Extraversion	0.133333	0.060000	0.106667	0.123333	0.126667	0.120000	0.090000	0.136667	0.043333	0.073333	0.050000
4	Agreeableness	-0.036667	-0.193333	-0.133333	-0.133333	-0.090000	0.046667	-0.016667	0.240000	0.160000	0.120000	0.083333
5	Non-Neuroticism	0.016667	-0.006667	-0.010000	-0.006667	-0.033333	0.046667	-0.023333	0.093333	0.046667	0.046667	-0.040000

[11]:

_b5._priority_calculation(
    correlation_coefficients = df_correlation_coefficients,
    col_name_ocean = 'Trait',
    threshold = 0.55,
    number_priority = 3,
    number_importance_traits = 3,
    out = False
)

_b5._save_logs(df = _b5.df_files_priority_, name = 'auto_characteristics_priorities_mupta_ru', out = True)

# Опционно
df = _b5.df_files_priority_.rename(columns = {'Openness':'OPE', 'Conscientiousness':'CON', 'Extraversion': 'EXT', 'Agreeableness': 'AGR', 'Non-Neuroticism': 'NNEU'})
columns_to_round = ['OPE', 'CON', 'EXT', 'AGR', 'NNEU']
df[columns_to_round] = df[columns_to_round].apply(lambda x: [round(i, 3) for i in x])
df

[11]:

	Path	OPE	CON	EXT	AGR	NNEU	Priority 1	Priority 2	Priority 3	Trait importance 1	Trait importance 2	Trait importance 3
Person ID
1	speaker_01_center_83.mov	0.766	0.697	0.656	0.760	0.494	Safe and reliable	Practical and easy to use	Economical/low cost	Conscientiousness	Agreeableness	Openness
2	speaker_06_center_83.mov	0.687	0.659	0.612	0.750	0.421	Safe and reliable	Practical and easy to use	Economical/low cost	Agreeableness	Conscientiousness	Openness
3	speaker_07_center_83.mov	0.672	0.661	0.572	0.705	0.381	Safe and reliable	Practical and easy to use	Economical/low cost	Conscientiousness	Agreeableness	Openness
4	speaker_10_center_83.mov	0.698	0.599	0.572	0.675	0.351	Safe and reliable	Practical and easy to use	Economical/low cost	Conscientiousness	Agreeableness	Openness
5	speaker_11_center_83.mov	0.718	0.599	0.574	0.732	0.380	Safe and reliable	Practical and easy to use	Economical/low cost	Agreeableness	Conscientiousness	Openness
6	speaker_15_center_83.mov	0.671	0.671	0.602	0.709	0.400	Safe and reliable	Practical and easy to use	Economical/low cost	Conscientiousness	Agreeableness	Openness
7	speaker_19_center_83.mov	0.767	0.658	0.653	0.801	0.463	Safe and reliable	Practical and easy to use	Economical/low cost	Agreeableness	Conscientiousness	Openness
8	speaker_23_center_83.mov	0.700	0.685	0.617	0.806	0.448	Safe and reliable	Practical and easy to use	Economical/low cost	Agreeableness	Conscientiousness	Openness
9	speaker_24_center_83.mov	0.711	0.663	0.611	0.711	0.414	Safe and reliable	Practical and easy to use	Economical/low cost	Conscientiousness	Agreeableness	Openness
10	speaker_27_center_83.mov	0.759	0.713	0.658	0.831	0.508	Safe and reliable	Practical and easy to use	Economical/low cost	Agreeableness	Conscientiousness	Openness

Прогнозирование потребительских предпочтений на характеристики мобильного устройства

В качестве примера предлагается использование коэффициентов корреляции между персональными качествами человека и характеристиками мобильного устройства, представленными в статье:

Peltonen E., Sharmila P., Asare K. O., Visuri A., Lagerspetz E., Ferreira D. (2020). When phones get personal: Predicting Big Five personality traits from application usage // Pervasive and Mobile Computing. – 2020. – Vol. 69. – 101269.

[12]:

# Загрузка датафрейма с коэффициентами корреляции
url = 'https://download.sberdisk.ru/download/file/478676690?token=7KcAxPqMpWiYQnx&filename=divice_characteristics.csv'
df_divice_characteristics = pd.read_csv(url)

df_divice_characteristics.index.name = 'ID'
df_divice_characteristics.index += 1
df_divice_characteristics.index = df_divice_characteristics.index.map(str)

df_divice_characteristics

[12]:

	Trait	Communication	Game Action	Game Board	Game Casino	Game Educational	Game Simulation	Game Trivia	Entertainment	Finance	Health and Fitness	Media and Video	Music and Audio	News and Magazines	Personalisation	Travel and Local	Weather
ID
1	Openness	0.118	0.056	0.079	0.342	0.027	0.104	0.026	0.000	0.006	0.002	0.000	0.000	0.001	0.004	0.002	0.004
2	Conscientiousness	0.119	0.043	0.107	0.448	0.039	0.012	0.119	0.000	0.005	0.001	0.000	0.002	0.002	0.001	0.001	0.003
3	Extraversion	0.246	0.182	0.211	0.311	0.102	0.165	0.223	0.001	0.003	0.000	0.001	0.001	0.001	0.004	0.009	0.003
4	Agreeableness	0.218	0.104	0.164	0.284	0.165	0.122	0.162	0.000	0.003	0.001	0.000	0.002	0.002	0.001	0.004	0.003
5	Non-Neuroticism	0.046	0.047	0.125	0.515	0.272	0.179	0.214	0.002	0.030	0.001	0.000	0.005	0.003	0.008	0.004	0.007

[13]:

_b5._priority_calculation(
    correlation_coefficients = df_divice_characteristics,
    col_name_ocean = 'Trait',
    threshold = 0.55,
    number_priority = 3,
    number_importance_traits = 3,
    out = True
)

_b5._save_logs(df = _b5.df_files_priority_, name = 'divice_characteristics_priorities_mupta_ru', out = True)

# Опционно
df = _b5.df_files_priority_.rename(columns = {'Openness':'OPE', 'Conscientiousness':'CON', 'Extraversion': 'EXT', 'Agreeableness': 'AGR', 'Non-Neuroticism': 'NNEU'})
columns_to_round = ['OPE', 'CON', 'EXT', 'AGR', 'NNEU']
df[columns_to_round] = df[columns_to_round].apply(lambda x: [round(i, 3) for i in x])
df

[13]:

	Path	OPE	CON	EXT	AGR	NNEU	Priority 1	Priority 2	Priority 3	Trait importance 1	Trait importance 2	Trait importance 3
Person ID
1	speaker_01_center_83.mov	0.766	0.697	0.656	0.760	0.494	Game Casino	Communication	Game Board	Agreeableness	Extraversion	Conscientiousness
2	speaker_06_center_83.mov	0.687	0.659	0.612	0.750	0.421	Game Casino	Communication	Game Board	Agreeableness	Extraversion	Conscientiousness
3	speaker_07_center_83.mov	0.672	0.661	0.572	0.705	0.381	Game Casino	Communication	Game Board	Agreeableness	Conscientiousness	Extraversion
4	speaker_10_center_83.mov	0.698	0.599	0.572	0.675	0.351	Game Casino	Communication	Game Board	Agreeableness	Extraversion	Conscientiousness
5	speaker_11_center_83.mov	0.718	0.599	0.574	0.732	0.380	Game Casino	Communication	Game Board	Agreeableness	Extraversion	Conscientiousness
6	speaker_15_center_83.mov	0.671	0.671	0.602	0.709	0.400	Game Casino	Communication	Game Board	Agreeableness	Extraversion	Conscientiousness
7	speaker_19_center_83.mov	0.767	0.658	0.653	0.801	0.463	Game Casino	Communication	Game Board	Agreeableness	Extraversion	Conscientiousness
8	speaker_23_center_83.mov	0.700	0.685	0.617	0.806	0.448	Game Casino	Communication	Game Board	Agreeableness	Extraversion	Conscientiousness
9	speaker_24_center_83.mov	0.711	0.663	0.611	0.711	0.414	Game Casino	Communication	Game Board	Agreeableness	Extraversion	Conscientiousness
10	speaker_27_center_83.mov	0.759	0.713	0.658	0.831	0.508	Game Casino	Communication	Game Board	Agreeableness	Extraversion	Conscientiousness

Прогнозирование потребительских предпочтений по стилю одежды

В качестве примера предлагается использование коэффициентов корреляции между персональными качествами человека и стилем одежды, представленными в статье:

Stolovy T. Styling the self: clothing practices, personality traits, and body image among Israeli women // Frontiers in psychology. - 2022. - vol. 12. - 719318.

[14]:

# Загрузка датафрейма с коэффициентами корреляции
url = 'https://download.sberdisk.ru/download/file/493644097?token=KGtSGMxjZtWXmBz&filename=df_%D1%81lothing_style_correlation.csv'
df_clothing_styles = pd.read_csv(url)

df_clothing_styles.index.name = 'ID'
df_clothing_styles.index += 1
df_clothing_styles.index = df_clothing_styles.index.map(str)

df_clothing_styles

[14]:

	Trait	Comfort	Camouflage	Assurance	Fashion	Individuality
ID
1	Openness	0.01	-0.24	0.31	0.07	0.31
2	Conscientiousness	-0.03	-0.24	0.17	0.09	0.15
3	Extraversion	-0.01	-0.19	0.30	0.13	0.14
4	Agreeableness	0.16	-0.16	0.15	-0.09	-0.05
5	Non-Neuroticism	0.03	-0.16	0.01	0.00	0.06

[15]:

_b5._priority_calculation(
    correlation_coefficients = df_clothing_styles,
    col_name_ocean = 'Trait',
    threshold = 0.55,
    number_priority = 3,
    number_importance_traits = 3,
    out = True
)

_b5._save_logs(df = _b5.df_files_priority_, name = 'clothing_styles_priorities_mupta_ru', out = True)

# Опционно
df = _b5.df_files_priority_.rename(columns = {'Openness':'OPE', 'Conscientiousness':'CON', 'Extraversion': 'EXT', 'Agreeableness': 'AGR', 'Non-Neuroticism': 'NNEU'})
columns_to_round = ['OPE', 'CON', 'EXT', 'AGR', 'NNEU']
df[columns_to_round] = df[columns_to_round].apply(lambda x: [round(i, 3) for i in x])
df

[15]:

	Path	OPE	CON	EXT	AGR	NNEU	Priority 1	Priority 2	Priority 3	Trait importance 1	Trait importance 2	Trait importance 3
Person ID
1	speaker_01_center_83.mov	0.766	0.697	0.656	0.760	0.494	Assurance	Individuality	Fashion	Openness	Extraversion	Conscientiousness
2	speaker_06_center_83.mov	0.687	0.659	0.612	0.750	0.421	Assurance	Individuality	Fashion	Openness	Extraversion	Conscientiousness
3	speaker_07_center_83.mov	0.672	0.661	0.572	0.705	0.381	Assurance	Individuality	Fashion	Openness	Extraversion	Conscientiousness
4	speaker_10_center_83.mov	0.698	0.599	0.572	0.675	0.351	Assurance	Individuality	Fashion	Openness	Extraversion	Conscientiousness
5	speaker_11_center_83.mov	0.718	0.599	0.574	0.732	0.380	Assurance	Individuality	Fashion	Openness	Extraversion	Conscientiousness
6	speaker_15_center_83.mov	0.671	0.671	0.602	0.709	0.400	Assurance	Individuality	Fashion	Openness	Extraversion	Conscientiousness
7	speaker_19_center_83.mov	0.767	0.658	0.653	0.801	0.463	Assurance	Individuality	Fashion	Openness	Extraversion	Conscientiousness
8	speaker_23_center_83.mov	0.700	0.685	0.617	0.806	0.448	Assurance	Individuality	Fashion	Openness	Extraversion	Conscientiousness
9	speaker_24_center_83.mov	0.711	0.663	0.611	0.711	0.414	Assurance	Individuality	Fashion	Openness	Extraversion	Conscientiousness
10	speaker_27_center_83.mov	0.759	0.713	0.658	0.831	0.508	Assurance	Individuality	Fashion	Openness	Extraversion	Conscientiousness

`MuPTA` (en)

[16]:

import os
import pandas as pd

# Импорт модуля
from oceanai.modules.lab.build import Run

# Создание экземпляра класса
_b5 = Run()

corpus = 'fi'
lang = 'en'

# Настройка ядра
_b5.path_to_save_ = './models' # Директория для сохранения файла
_b5.chunk_size_ = 2000000      # Размер загрузки файла из сети за 1 шаг

# Формирование аудиомоделей
res_load_model_hc = _b5.load_audio_model_hc()
res_load_model_nn = _b5.load_audio_model_nn()

# Загрузка весов аудиомоделей
url = _b5.weights_for_big5_['audio'][corpus]['hc']['googledisk']
res_load_model_weights_hc = _b5.load_audio_model_weights_hc(url = url, force_reload = False)

url = _b5.weights_for_big5_['audio'][corpus]['nn']['googledisk']
res_load_model_weights_nn = _b5.load_audio_model_weights_nn(url = url, force_reload = False)

# Формирование видеомоделей
res_load_model_hc = _b5.load_video_model_hc(lang=lang)
res_load_model_deep_fe = _b5.load_video_model_deep_fe()
res_load_model_nn = _b5.load_video_model_nn()

# Загрузка весов видеомоделей
url = _b5.weights_for_big5_['video'][corpus]['hc']['googledisk']
res_load_model_weights_hc = _b5.load_video_model_weights_hc(url = url, force_reload = False)

url = _b5.weights_for_big5_['video'][corpus]['fe']['googledisk']
res_load_model_weights_deep_fe = _b5.load_video_model_weights_deep_fe(url = url, force_reload = False)

url = _b5.weights_for_big5_['video'][corpus]['nn']['googledisk']
res_load_model_weights_nn = _b5.load_video_model_weights_nn(url = url, force_reload = False)

# Загрузка словаря с экспертными признаками (текстовая модальность)
res_load_text_features = _b5.load_text_features()

# Формирование текстовых моделей
res_setup_translation_model = _b5.setup_translation_model() # только для русского языка
res_setup_translation_model = _b5.setup_bert_encoder(force_reload = False)
res_load_text_model_hc_fi = _b5.load_text_model_hc(corpus=corpus)
res_load_text_model_nn_fi = _b5.load_text_model_nn(corpus=corpus)

# Загрузка весов текстовых моделей
url = _b5.weights_for_big5_['text'][corpus]['hc']['googledisk']
res_load_text_model_weights_hc_fi = _b5.load_text_model_weights_hc(url = url, force_reload = False)

url = _b5.weights_for_big5_['text'][corpus]['nn']['googledisk']
res_load_text_model_weights_nn_fi = _b5.load_text_model_weights_nn(url = url, force_reload = False)

# Формирование модели для мультимодального объединения информации
res_load_avt_model_b5 = _b5.load_avt_model_b5()

# Загрузка весов модели для мультимодального объединения информации
url = _b5.weights_for_big5_['avt'][corpus]['b5']['googledisk']
res_load_avt_model_weights_b5 = _b5.load_avt_model_weights_b5(url = url, force_reload = False)

PATH_TO_DIR = './video_MuPTA/'
PATH_SAVE_VIDEO = './video_MuPTA/test/'

_b5.path_to_save_ = PATH_SAVE_VIDEO

# Загрузка 10 тестовых аудиовидеозаписей из корпуса MuPTA
# URL: https://hci.nw.ru/en/pages/mupta-corpus
domain = 'https://download.sberdisk.ru/download/file/'
tets_name_files = [
    '477995979?token=2cvyk7CS0mHx2MJ&filename=speaker_06_center_83.mov',
    '477995980?token=jGPtBPS69uzFU6Y&filename=speaker_01_center_83.mov',
    '477995967?token=zCaRbNB6ht5wMPq&filename=speaker_11_center_83.mov',
    '477995966?token=B1rbinDYRQKrI3T&filename=speaker_15_center_83.mov',
    '477995978?token=dEpVDtZg1EQiEQ9&filename=speaker_07_center_83.mov',
    '477995961?token=o1hVjw8G45q9L9Z&filename=speaker_19_center_83.mov',
    '477995964?token=5K220Aqf673VHPq&filename=speaker_23_center_83.mov',
    '477995965?token=v1LVD2KT1cU7Lpb&filename=speaker_24_center_83.mov',
    '477995962?token=tmaSGyyWLA6XCy9&filename=speaker_27_center_83.mov',
    '477995963?token=bTpo96qNDPcwGqb&filename=speaker_10_center_83.mov',
]

for curr_files in tets_name_files:
    _b5.download_file_from_url(url = domain + curr_files, out = True)

# Получение прогнозов
_b5.path_to_dataset_ = PATH_TO_DIR # Директория набора данных
_b5.ext_ = ['.mov'] # Расширения искомых файлов

# Полный путь к файлу с верными предсказаниями для подсчета точности
url_accuracy = _b5.true_traits_['mupta']['sberdisk']

_b5.get_avt_predictions(url_accuracy = url_accuracy, lang = lang)

[2024-10-10 18:29:55] Извлечение признаков (экспертных и нейросетевых) из текста …

[2024-10-10 18:29:56] Получение прогнозов и вычисление точности (мультимодальное объединение) …

10 из 10 (100.0%) … GitHub:nbsphinx-math:OCEANAI\docs\source\user_guide:nbsphinx-math:notebooks\video_MuPTA:nbsphinx-math:test\speaker_27_center_83.mov …

	Path	Openness	Conscientiousness	Extraversion	Agreeableness	Non-Neuroticism
Person ID
1	speaker_01_center_83.mov	0.59561	0.542967	0.440668	0.589769	0.515306
2	speaker_06_center_83.mov	0.661347	0.673973	0.603208	0.64543	0.6431
3	speaker_07_center_83.mov	0.439868	0.465049	0.284547	0.422551	0.396058
4	speaker_10_center_83.mov	0.47715	0.502563	0.373686	0.441372	0.424637
5	speaker_11_center_83.mov	0.403292	0.344359	0.317304	0.422228	0.384346
6	speaker_15_center_83.mov	0.581837	0.562177	0.504623	0.602169	0.522254
7	speaker_19_center_83.mov	0.510444	0.448468	0.425599	0.451861	0.447891
8	speaker_23_center_83.mov	0.500526	0.541376	0.308529	0.441178	0.452412
9	speaker_24_center_83.mov	0.427677	0.511355	0.301078	0.434281	0.442301
10	speaker_27_center_83.mov	0.566414	0.659169	0.434059	0.59122	0.579172

[2024-10-10 18:29:56] Точность по отдельным персональным качествам личности человека …

	Openness	Conscientiousness	Extraversion	Agreeableness	Non-Neuroticism	Mean
Metrics
MAE	0.1632	0.1621	0.176	0.2589	0.1122	0.1745
Accuracy	0.8368	0.8379	0.824	0.7411	0.8878	0.8255

[2024-10-10 18:29:56] Средняя средних абсолютных ошибок: 0.1745, средняя точность: 0.8255 …

Лог файлы успешно сохранены …

— Время выполнения: 320.737 сек. —

[16]:

True

Для прогнозирования потребительских предпочтений в промышленных товарах необходимо знать коэффициенты корреляции, определяющие взаимосвязь между персональными качествами личности человека и предпочтениями в товарах или услугах.

В качестве примера предлагается использование коэффициентов корреляции между персональными качествами человека и характеристиками автомобилей, представленными в статье:

O’Connor P. J. et al. What Drives Consumer Automobile Choice? Investigating Personality Trait Predictors of Vehicle Preference Factors // Personality and Individual Differences. – 2022. – Vol. 184. – pp. 111220.

Пользователь может установить свои коэффициенты корреляции.

Прогнозирование потребительских предпочтений на характеристики атомобиля

[17]:

# Загрузка датафрейма с коэффициентами корреляции
url = 'https://download.sberdisk.ru/download/file/478675818?token=EjfLMqOeK8cfnOu&filename=auto_characteristics.csv'
df_correlation_coefficients = pd.read_csv(url)
df_correlation_coefficients = pd.DataFrame(
    df_correlation_coefficients.drop(['Style and performance', 'Safety and practicality'], axis = 1)
)
df_correlation_coefficients.index.name = 'ID'
df_correlation_coefficients.index += 1
df_correlation_coefficients.index = df_correlation_coefficients.index.map(str)

df_correlation_coefficients

[17]:

	Trait	Performance	Classic car features	Luxury additions	Fashion and attention	Recreation	Technology	Family friendly	Safe and reliable	Practical and easy to use	Economical/low cost	Basic features
ID
1	Openness	0.020000	-0.033333	-0.030000	-0.050000	0.033333	0.013333	-0.030000	0.136667	0.106667	0.093333	0.006667
2	Conscientiousness	0.013333	-0.193333	-0.063333	-0.096667	-0.096667	0.086667	-0.063333	0.280000	0.180000	0.130000	0.143333
3	Extraversion	0.133333	0.060000	0.106667	0.123333	0.126667	0.120000	0.090000	0.136667	0.043333	0.073333	0.050000
4	Agreeableness	-0.036667	-0.193333	-0.133333	-0.133333	-0.090000	0.046667	-0.016667	0.240000	0.160000	0.120000	0.083333
5	Non-Neuroticism	0.016667	-0.006667	-0.010000	-0.006667	-0.033333	0.046667	-0.023333	0.093333	0.046667	0.046667	-0.040000

[18]:

_b5._priority_calculation(
    correlation_coefficients = df_correlation_coefficients,
    col_name_ocean = 'Trait',
    threshold = 0.55,
    number_priority = 3,
    number_importance_traits = 3,
    out = False
)

_b5._save_logs(df = _b5.df_files_priority_, name = 'auto_characteristics_priorities_mupta_en', out = True)

# Опционно
df = _b5.df_files_priority_.rename(columns = {'Openness':'OPE', 'Conscientiousness':'CON', 'Extraversion': 'EXT', 'Agreeableness': 'AGR', 'Non-Neuroticism': 'NNEU'})
columns_to_round = ['OPE', 'CON', 'EXT', 'AGR', 'NNEU']
df[columns_to_round] = df[columns_to_round].apply(lambda x: [round(i, 3) for i in x])
df

[18]:

	Path	OPE	CON	EXT	AGR	NNEU	Priority 1	Priority 2	Priority 3	Trait importance 1	Trait importance 2	Trait importance 3
Person ID
1	speaker_01_center_83.mov	0.596	0.543	0.441	0.590	0.515	Practical and easy to use	Economical/low cost	Recreation	Openness	Agreeableness	Non-Neuroticism
2	speaker_06_center_83.mov	0.661	0.674	0.603	0.645	0.643	Safe and reliable	Practical and easy to use	Economical/low cost	Conscientiousness	Agreeableness	Openness
3	speaker_07_center_83.mov	0.440	0.465	0.285	0.423	0.396	Classic car features	Fashion and attention	Luxury additions	Agreeableness	Conscientiousness	Openness
4	speaker_10_center_83.mov	0.477	0.503	0.374	0.441	0.425	Classic car features	Fashion and attention	Luxury additions	Agreeableness	Conscientiousness	Openness
5	speaker_11_center_83.mov	0.403	0.344	0.317	0.422	0.384	Classic car features	Fashion and attention	Luxury additions	Agreeableness	Conscientiousness	Openness
6	speaker_15_center_83.mov	0.582	0.562	0.505	0.602	0.522	Safe and reliable	Practical and easy to use	Economical/low cost	Conscientiousness	Agreeableness	Openness
7	speaker_19_center_83.mov	0.510	0.448	0.426	0.452	0.448	Classic car features	Fashion and attention	Luxury additions	Agreeableness	Conscientiousness	Openness
8	speaker_23_center_83.mov	0.501	0.541	0.309	0.441	0.452	Classic car features	Fashion and attention	Luxury additions	Agreeableness	Conscientiousness	Openness
9	speaker_24_center_83.mov	0.428	0.511	0.301	0.434	0.442	Classic car features	Fashion and attention	Luxury additions	Agreeableness	Conscientiousness	Openness
10	speaker_27_center_83.mov	0.566	0.659	0.434	0.591	0.579	Safe and reliable	Practical and easy to use	Economical/low cost	Conscientiousness	Agreeableness	Openness

Прогнозирование потребительских предпочтений на характеристики мобильного устройства

В качестве примера предлагается использование коэффициентов корреляции между персональными качествами человека и характеристиками мобильного устройства, представленными в статье:

Peltonen E., Sharmila P., Asare K. O., Visuri A., Lagerspetz E., Ferreira D. (2020). When phones get personal: Predicting Big Five personality traits from application usage // Pervasive and Mobile Computing. – 2020. – Vol. 69. – 101269.

[19]:

# Загрузка датафрейма с коэффициентами корреляции
url = 'https://download.sberdisk.ru/download/file/478676690?token=7KcAxPqMpWiYQnx&filename=divice_characteristics.csv'
df_divice_characteristics = pd.read_csv(url)

df_divice_characteristics.index.name = 'ID'
df_divice_characteristics.index += 1
df_divice_characteristics.index = df_divice_characteristics.index.map(str)

df_divice_characteristics

[19]:

	Trait	Communication	Game Action	Game Board	Game Casino	Game Educational	Game Simulation	Game Trivia	Entertainment	Finance	Health and Fitness	Media and Video	Music and Audio	News and Magazines	Personalisation	Travel and Local	Weather
ID
1	Openness	0.118	0.056	0.079	0.342	0.027	0.104	0.026	0.000	0.006	0.002	0.000	0.000	0.001	0.004	0.002	0.004
2	Conscientiousness	0.119	0.043	0.107	0.448	0.039	0.012	0.119	0.000	0.005	0.001	0.000	0.002	0.002	0.001	0.001	0.003
3	Extraversion	0.246	0.182	0.211	0.311	0.102	0.165	0.223	0.001	0.003	0.000	0.001	0.001	0.001	0.004	0.009	0.003
4	Agreeableness	0.218	0.104	0.164	0.284	0.165	0.122	0.162	0.000	0.003	0.001	0.000	0.002	0.002	0.001	0.004	0.003
5	Non-Neuroticism	0.046	0.047	0.125	0.515	0.272	0.179	0.214	0.002	0.030	0.001	0.000	0.005	0.003	0.008	0.004	0.007

[20]:

_b5._priority_calculation(
    correlation_coefficients = df_divice_characteristics,
    col_name_ocean = 'Trait',
    threshold = 0.55,
    number_priority = 3,
    number_importance_traits = 3,
    out = True
)

_b5._save_logs(df = _b5.df_files_priority_, name = 'divice_characteristics_priorities_mupta_en', out = True)

# Опционно
df = _b5.df_files_priority_.rename(columns = {'Openness':'OPE', 'Conscientiousness':'CON', 'Extraversion': 'EXT', 'Agreeableness': 'AGR', 'Non-Neuroticism': 'NNEU'})
columns_to_round = ['OPE', 'CON', 'EXT', 'AGR', 'NNEU']
df[columns_to_round] = df[columns_to_round].apply(lambda x: [round(i, 3) for i in x])
df

[20]:

	Path	OPE	CON	EXT	AGR	NNEU	Priority 1	Priority 2	Priority 3	Trait importance 1	Trait importance 2	Trait importance 3
Person ID
1	speaker_01_center_83.mov	0.596	0.543	0.441	0.590	0.515	Communication	Health and Fitness	Media and Video	Agreeableness	Openness	Non-Neuroticism
2	speaker_06_center_83.mov	0.661	0.674	0.603	0.645	0.643	Game Casino	Communication	Game Trivia	Non-Neuroticism	Extraversion	Conscientiousness
3	speaker_07_center_83.mov	0.440	0.465	0.285	0.423	0.396	Media and Video	Entertainment	Health and Fitness	Agreeableness	Conscientiousness	Extraversion
4	speaker_10_center_83.mov	0.477	0.503	0.374	0.441	0.425	Media and Video	Entertainment	Health and Fitness	Agreeableness	Conscientiousness	Extraversion
5	speaker_11_center_83.mov	0.403	0.344	0.317	0.422	0.384	Media and Video	Entertainment	Health and Fitness	Conscientiousness	Agreeableness	Extraversion
6	speaker_15_center_83.mov	0.582	0.562	0.505	0.602	0.522	Game Casino	Communication	Game Board	Agreeableness	Conscientiousness	Openness
7	speaker_19_center_83.mov	0.510	0.448	0.426	0.452	0.448	Media and Video	Entertainment	Health and Fitness	Conscientiousness	Agreeableness	Extraversion
8	speaker_23_center_83.mov	0.501	0.541	0.309	0.441	0.452	Media and Video	Entertainment	Health and Fitness	Agreeableness	Conscientiousness	Extraversion
9	speaker_24_center_83.mov	0.428	0.511	0.301	0.434	0.442	Media and Video	Entertainment	Health and Fitness	Agreeableness	Conscientiousness	Extraversion
10	speaker_27_center_83.mov	0.566	0.659	0.434	0.591	0.579	Game Casino	Game Educational	Game Trivia	Non-Neuroticism	Conscientiousness	Agreeableness

Прогнозирование потребительских предпочтений по стилю одежды

В качестве примера предлагается использование коэффициентов корреляции между персональными качествами человека и стилем одежды, представленными в статье:

Stolovy T. Styling the self: clothing practices, personality traits, and body image among Israeli women // Frontiers in psychology. - 2022. - vol. 12. - 719318.

[21]:

# Загрузка датафрейма с коэффициентами корреляции
url = 'https://download.sberdisk.ru/download/file/493644097?token=KGtSGMxjZtWXmBz&filename=df_%D1%81lothing_style_correlation.csv'
df_clothing_styles = pd.read_csv(url)

df_clothing_styles.index.name = 'ID'
df_clothing_styles.index += 1
df_clothing_styles.index = df_clothing_styles.index.map(str)

df_clothing_styles

[21]:

	Trait	Comfort	Camouflage	Assurance	Fashion	Individuality
ID
1	Openness	0.01	-0.24	0.31	0.07	0.31
2	Conscientiousness	-0.03	-0.24	0.17	0.09	0.15
3	Extraversion	-0.01	-0.19	0.30	0.13	0.14
4	Agreeableness	0.16	-0.16	0.15	-0.09	-0.05
5	Non-Neuroticism	0.03	-0.16	0.01	0.00	0.06

[22]:

_b5._priority_calculation(
    correlation_coefficients = df_clothing_styles,
    col_name_ocean = 'Trait',
    threshold = 0.55,
    number_priority = 3,
    number_importance_traits = 3,
    out = True
)

_b5._save_logs(df = _b5.df_files_priority_, name = 'clothing_styles_priorities_mupta_en', out = True)

# Опционно
df = _b5.df_files_priority_.rename(columns = {'Openness':'OPE', 'Conscientiousness':'CON', 'Extraversion': 'EXT', 'Agreeableness': 'AGR', 'Non-Neuroticism': 'NNEU'})
columns_to_round = ['OPE', 'CON', 'EXT', 'AGR', 'NNEU']
df[columns_to_round] = df[columns_to_round].apply(lambda x: [round(i, 3) for i in x])
df

[22]:

	Path	OPE	CON	EXT	AGR	NNEU	Priority 1	Priority 2	Priority 3	Trait importance 1	Trait importance 2	Trait importance 3
Person ID
1	speaker_01_center_83.mov	0.596	0.543	0.441	0.590	0.515	Comfort	Camouflage	Assurance	Agreeableness	Non-Neuroticism	Conscientiousness
2	speaker_06_center_83.mov	0.661	0.674	0.603	0.645	0.643	Assurance	Individuality	Fashion	Openness	Extraversion	Conscientiousness
3	speaker_07_center_83.mov	0.440	0.465	0.285	0.423	0.396	Camouflage	Comfort	Fashion	Conscientiousness	Openness	Non-Neuroticism
4	speaker_10_center_83.mov	0.477	0.503	0.374	0.441	0.425	Camouflage	Comfort	Fashion	Conscientiousness	Openness	Non-Neuroticism
5	speaker_11_center_83.mov	0.403	0.344	0.317	0.422	0.384	Camouflage	Fashion	Comfort	Openness	Conscientiousness	Non-Neuroticism
6	speaker_15_center_83.mov	0.582	0.562	0.505	0.602	0.522	Assurance	Individuality	Comfort	Openness	Conscientiousness	Agreeableness
7	speaker_19_center_83.mov	0.510	0.448	0.426	0.452	0.448	Camouflage	Comfort	Fashion	Openness	Conscientiousness	Non-Neuroticism
8	speaker_23_center_83.mov	0.501	0.541	0.309	0.441	0.452	Camouflage	Comfort	Fashion	Conscientiousness	Openness	Non-Neuroticism
9	speaker_24_center_83.mov	0.428	0.511	0.301	0.434	0.442	Camouflage	Comfort	Fashion	Conscientiousness	Openness	Non-Neuroticism
10	speaker_27_center_83.mov	0.566	0.659	0.434	0.591	0.579	Assurance	Individuality	Comfort	Openness	Conscientiousness	Agreeableness

Решение практической задачи 2

Задача: прогнозирование потребительских предпочтений на промышленные товары

FI V2

Прогнозирование потребительских предпочтений на характеристики атомобиля

Прогнозирование потребительских предпочтений на характеристики мобильного устройства

Прогнозирование потребительских предпочтений по стилю одежды

MuPTA (ru)

Прогнозирование потребительских предпочтений на характеристики атомобиля

Прогнозирование потребительских предпочтений на характеристики мобильного устройства

Прогнозирование потребительских предпочтений по стилю одежды

MuPTA (en)

Прогнозирование потребительских предпочтений на характеристики атомобиля

Прогнозирование потребительских предпочтений на характеристики мобильного устройства

Прогнозирование потребительских предпочтений по стилю одежды

`FI V2`

`MuPTA` (ru)

`MuPTA` (en)