The bank wants to explore ways of converting its liability customers to personal loan customers (while retaining them as depositors). A campaign that the bank ran last year for liability customers showed a conversion rate of over 9% success. This has encouraged the retail marketing department to devise campaigns with better target marketing to increase the success ratio with minimal budget.

• Our Goal
• Presentation File
• Notebook Files
• How Were We Able To Predict The Progression
  • Feature Engineering
  • Handling Missing Features
  • Scaling
  • Feature Selection
  • Handling Imbalanced Data
  • Model Evaluation and Selection
• Our Primary Target Audience
• Coding Scripts
• Web Application
• Dockerfile

Using the Data set, help to build a Machine Learning Model which can predict if a Customer would take up Personal Loan if the targeted Marketing Campaign is done

Here is the link to the Presentation PDF

Here is the link to the EDA Notebook

Here is the link to the Feature Selection & Modelling

Explored various models like Random Forest, Gradient Boosting, XGBClassifier, KNeighborsClassifier using different hyperparameters

• Feature Engineered Experience :
  • Featured engineered Experience feature to gain insights with the data but the Heat map of Correlation matrix showed Experience field is not a good feature to conside

• Categorical data :
  • Replace missing values with mode
• Numerical Data :
  • Replace missing values with mean

• Performed Standard Scaling on the features

Performed correlation matrix to check which features are the best

I used SMOTE to improve the imbalanced data

The following were the performance metrics of the models

• Income ranging between 122k to 172k
• Credit card average spending ranging between 2.6k to 5.4k
• People with deposit account

1. data_ingestion.py : For ingesting train, test and raw data along with the trained model
2. data_transformation.py : For handling missing values, scaling and oversampling with SMOTE
3. model_trainer.py : For training different models on various hyper parameters and choosing the best based on ROC AUC Score
4. predict_pipeline.py : To predict the data created using CustomData class
5. app.py : For running the front end application to predict if the person will take personal loan

Other Scripts include :

1. utils.py : For common function used in the project
2. logger.py : For improved logging
3. exception.py : For logging errors along with line number of error along with file name

First clone the repo

git clone https://github.com/princyiakov/customer_conversion.git
cd customer_conversion

If you have docker installed, run the following command to access the webapp on http:https://localhost:3000

docker build -t customer_conversion .
docker run -p 3000:3000 customer_conversion