You can deploy this model and start earning money today!

joyjitroy/
Bank_Churn_Prediction

keras

finance

mit

Views

No views yet

💰 Bank Churn Prediction — AI for Smarter Customer Retention

🧩 Overview

Businesses like banks which provide service have to worry about the problem of Customer Churn, i.e. customers leaving and joining another service provider. It is important to understand which aspects of the service influence a customer's decision in this regard.
Management can concentrate efforts on improvement of service, keeping in mind these priorities.

Objective
You as a Data Scientist with the bank need to build a neural network–based classifier that can determine whether a customer will leave the bank or not in the next 6 months.

🤖 Model Details

Model Type: Feed-forward ANN (Artificial Neural Network) — Binary Classifier
Framework: TensorFlow / Keras
Dataset: Churn.csv (Bank Customer Churn Dataset, 10,000+ customers)
Input: Structured customer profile (credit score, age, balance, tenure, activity, salary, etc.)
Output: Binary churn prediction (0 = stays, 1 = churn)

📊 Data Dictionary

Feature	Description
`CustomerId`	Unique ID assigned to each customer
`Surname`	Customer’s last name
`CreditScore`	Customer credit history
`Geography`	Customer location
`Gender`	Gender of the customer
`Age`	Age of the customer
`Tenure`	Number of years with the bank
`NumOfProducts`	Number of bank products purchased
`Balance`	Account balance
`HasCrCard`	Whether the customer has a credit card
`EstimatedSalary`	Estimated salary
`IsActiveMember`	Whether the customer is an active member
`Exited`	Target label — 0: No (customer stays), 1: Yes (customer churns)

🌟 Why It Matters

✅ Detects early signs of potential churn
✅ Enables targeted retention strategies
✅ Improves customer engagement and loyalty
✅ Helps maximize profitability and reduce attrition rates

📘 Full Source Notebook:
The complete training and evaluation notebook is available on GitHub:
👉 View on GitHub

🚀 Example Usage

python

# Inference example aligned to your dataset schema

import pandas as pd
import numpy as np

CSV_PATH = "Data/Churn.csv"
TARGET = "Exited"

# Columns exactly as in your table
ALL_COLS = [
    "CustomerId","Surname","CreditScore","Geography","Gender","Age","Tenure",
    "NumOfProducts","Balance","HasCrCard","EstimatedSalary","IsActiveMember","Exited"
]

NUM_COLS = ["CreditScore","Age","Tenure","NumOfProducts","Balance","EstimatedSalary","HasCrCard","IsActiveMember"]
CAT_COLS = ["Geography","Gender"]
DROP_COLS = ["CustomerId","Surname"]

# Load data
df = pd.read_csv(CSV_PATH)[ALL_COLS]

def prepare_features(df_in, fit_cols=None):
    X = df_in.drop(columns=[TARGET] + DROP_COLS).copy()
    # one hot on the two categoricals
    X = pd.get_dummies(X, columns=CAT_COLS, drop_first=True)
    # align to training columns
    if fit_cols is not None:
        X = X.reindex(columns=fit_cols, fill_value=0)
    return X

# If you trained in this notebook, reuse `model` and `feature_cols` from training:
#   model.save("bank_churn_ann.keras")
#   feature_cols = X_train.columns.tolist()

# If loading a saved model:
# from tensorflow.keras.models import load_model
# model = load_model("bank_churn_ann.keras")
# feature_cols = [...]  # same list you used during training after get_dummies

# Single example constructed with your schema
example = {
    "CustomerId": 15788241,
    "Surname": "Smith",
    "CreditScore": 600,
    "Geography": "Germany",   # France, Germany, Spain in this dataset
    "Gender": "Male",         # Male or Female
    "Age": 40,
    "Tenure": 3,
    "NumOfProducts": 2,
    "Balance": 60000.0,
    "HasCrCard": 1,
    "EstimatedSalary": 50000.0,
    "IsActiveMember": 1,
    "Exited": 0               # ignored at inference
}

ex_df = pd.DataFrame([example])
X_ex = prepare_features(ex_df, fit_cols=feature_cols)

pred_prob = float(model.predict(X_ex, verbose=0)[0][0])
pred = int(pred_prob >= 0.5)

print(f"Churn Probability: {pred_prob:.4f}")
print("Prediction:", "Likely to churn" if pred else "Likely to stay")

joyjitroy/Bank_Churn_Prediction