Input a customer. Get their segment, churn risk, and next month's spend — instantly.

---

🎯 What It Does

Three ML models. One dashboard. Actionable intelligence in seconds.

🔵 Segmentation K-Means Clustering Groups customers into 3 behavioral clusters for targeted marketing strategies

🔴 Churn Risk Logistic Regression Predicts High Risk vs Low Risk — act before the customer leaves

🟢 Future Value Linear Regression Forecasts exact $ spend next month for revenue planning

---

🔬 ML Pipeline — Deep Dive

  RAW INPUT (9 Features)
  Age · Gender · Location · Tenure · Avg Monthly Spend
  Last Month Spend · Num Transactions · Days Since Purchase · Support Tickets
          │
          ▼
  ┌───────────────────────┐
  │   DATA PREPROCESSING   │
  │  Label Encode Gender   │
  │  One-Hot Encode Location│
  │  StandardScaler        │
  └───────────────────────┘
          │
          ▼
  ┌───────────────────────┐
  │         PCA            │  → Dimensionality reduction
  │   (pca.pkl loaded)     │    Captures max variance
  └───────────────────────┘
          │
    ┌─────┴──────┬──────────────┐
    ▼            ▼              ▼
┌────────┐  ┌─────────┐  ┌──────────┐
│K-Means │  │Logistic │  │  Linear  │
│Cluster │  │Regress. │  │Regress.  │
│kmeans  │  │Classif. │  │Regress.  │
│.pkl    │  │_Model   │  │_Model    │
│        │  │.pkl     │  │.pkl      │
└────────┘  └─────────┘  └──────────┘
    │            │              │
    ▼            ▼              ▼
 Group 0/1/2  🔴 High Risk   💵 $XXX.XX
              🟢 Low Risk    next month

---

🖥️ App Workflow

1. OPEN DASHBOARD
   └── Sidebar: Choose input mode
         ├── 📋 Existing Customer  →  Select Customer ID
         │                             Auto-populates all fields from CSV
         └── ✏️  Manual Entry      →  Fill in 9 feature fields manually

2. HIT PREDICT
   └── Inputs encoded → scaled → PCA transformed
         └── Passed simultaneously to all 3 pkl models

3. VIEW RESULTS
   ├── 🔵 Customer Segment    →  Group 0 / 1 / 2
   ├── 🔴🟢 Churn Risk        →  High Risk (red) or Low Risk (green)
   └── 💵 Predicted Spend     →  $XXX.XX next month

---

🧪 Model Details

🔴 Churn Classification — Logistic Regression

• Tuned params: C, penalty, solver, class_weight
• Method: RandomizedSearchCV — 5-fold Cross-Validation
• Output: Binary — High Risk (1) or Low Risk (0)
• Metric: Accuracy, Precision, Recall, F1-score

🟢 Spend Regression — Linear Regression

• Tuned params: fit_intercept, positive
• Method: RandomizedSearchCV to minimize MAE & RMSE
• Output: Continuous — predicted $ spend next month
• Metric: MAE, RMSE

🔵 Customer Clustering — K-Means

• k = 3 — determined by Elbow Method (WCSS vs k plot)
• Validated with Silhouette Score
• Output: Cluster label — Group 0, 1, or 2
• Trained on PCA-reduced feature space

---

⚙️ Tech Stack

---

📁 Project Structure

Smart-Sales/
├── app.py                        # Streamlit UI + inference logic
├── Dataset/
│   ├── customer_data.csv         # Raw customer profiles
│   └── preprocessed_data.csv     # Cleaned & encoded training data
├── Models/
│   ├── Data_Preprocessing.py     # Cleaning, encoding, train/test split
│   ├── Classification_Model.py   # Trains churn logistic regression
│   ├── Regression_Model.py       # Trains spend linear regression
│   └── Unsupervised_model.py     # Trains K-Means with Elbow method
└── pkl/
    ├── scaler.pkl                 # StandardScaler artifact
    ├── pca.pkl                    # PCA model artifact
    ├── gender_encoder.pkl         # LabelEncoder for Gender
    ├── Classification_Model.pkl   # Churn prediction model
    ├── Regression_Model.pkl       # Spend forecast model
    └── kmeans_model.pkl           # Customer clustering model

---

🚀 Getting Started

# Clone the repo
git clone https://github.com/Ronit178693/Smart-Sales-Custom.git
cd Smart-Sales-Custom

# Create virtual environment
python -m venv venv
source venv/bin/activate       # Windows: venv\Scripts\activate

# Install dependencies
pip install -r requirements.txt

# Train models (generates .pkl files)
python Models/Data_Preprocessing.py
python Models/Classification_Model.py
python Models/Regression_Model.py
python Models/Unsupervised_model.py

# Launch the dashboard
streamlit run app.py

---

📈 Business Impact

Intelligence	Business Use
🔵 Customer Segments	Tailor campaigns per cluster — stop generic blasting
🔴 Churn Risk	Trigger retention offers before the customer leaves
💵 Future Spend	Forecast next month's revenue with customer-level precision
📊 Combined View	Prioritize high-value, low-churn-risk customers for upsells

---

Built with 🧠 by Ronit Agrawal