Skip to content

Building prediction model (ML) for whether the client will leave the bank soon using the real data from the bank X. The dataset is provided by Yandex Practicum.

Notifications You must be signed in to change notification settings

VeronicaCodes/ml_bank_churn_prediction

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

1 Commit
 
 
 
 
 
 

Repository files navigation

Bank customer churn prediction

Goals:

The goal was to predict whether a customer will leave the bank soon. The Bank had provided us with the historical data on clients’ past behavior and termination of contracts with the bank. It was necessary to build a model with the maximum possible F1 score. The bank's requirement for F1 score was it being equal to at least 0.59

Tools used:

pandas, numpy, sklearn, DecisionTreeClassifier, RandomForestClassifier, LogisticRegression, GradientBoostingClassifier

Conclusion:

model f1_valid AUC ROC
DecisionTree 0.56 0.82
RandomForest 0.60 0.86
LogisticRegression 0.45 0.71
GradientBoosting 0.60 0.83

'Random Forest Classifier' with max_depth=5 and n_estimators=10 is the best model, with f1 score equal to 0.50, and AUC ROC equal to 0.85.


Прогнозирование оттока клиентов банка

Описание и цели:

Клиенты банка постепенно уходят, и банк пришел к выводу, что дешевле сохранить существующих клиентов, чем привлекать новых. У нас есть данные о прошлом поведении клиентов и расторжении договоров с банком. Нужно спрогнозировать, скоро ли клиент покинет банк.

Инструменты: pandas, numpy, sklearn, DecisionTreeClassifier, RandomForestClassifier, LogisticRegression, GradientBoostingClassifier

Выводы:

model f1_valid AUC ROC
DecisionTree 0.56 0.82
RandomForest 0.60 0.86
LogisticRegression 0.45 0.71
GradientBoosting 0.60 0.83

Лучшая оценка - f1_score = 0.60, auc_roc = 0.85 с моделью RandomForest.

About

Building prediction model (ML) for whether the client will leave the bank soon using the real data from the bank X. The dataset is provided by Yandex Practicum.

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published