📊 Customer Satisfaction Prediction System

An end-to-end Data Analytics + Machine Learning project that analyzes customer support tickets and predicts customer satisfaction using AI.
This project combines EDA, visualization, ML modeling, and a professional interactive dashboard built with Streamlit.

---

📌 Project Overview

Customer satisfaction directly impacts customer retention and brand reputation.
This project predicts customer satisfaction levels using Machine Learning and visualizes insights through Power BI and a Python-based dashboard.

The project demonstrates a complete analytics lifecycle: Data → ML Model → Predictions → Business Dashboards

---

🎯 Objectives

• Analyze customer support ticket data
• Predict customer satisfaction ratings
• Identify customers at risk of dissatisfaction
• Monitor data quality and model performance
• Present insights using Power BI and Python dashboards

---

🗂 Project Structure

Customer-Satisfaction-Prediction/ │ ├── app/ │ └── app.py # Streamlit Dashboard │ ├── data/ │ ├── customer_support_tickets.csv │ └── eda_clean.csv │ ├── notebook/ │ ├── 01_EDA.ipynb # Data analysis & visualization │ └── 02_ML_Model.ipynb # Model training │ ├── outputs/ │ ├── figures/ # EDA graphs │ ├── ml_figures/ # Model performance plots │ ├── ml_metrics.csv │ └── predictions.csv │ ├── powerbi_dashboard/ | ├── Customer_Satisfaction_Dashboard.pbix │ ├── Overview.png │ ├── Tickets.png │ ├── Voice Of Customer.png | ├──Data health.png │ ├── ML Performance.png │ └── Prediction.png │ ├── check_model.py
├── requirements.txt
└── README.md

---

🧠 Machine Learning Workflow

1️⃣ Exploratory Data Analysis (EDA)

• Missing values analysis
• Ticket distribution and trends
• Satisfaction patterns by channel and priority
• Text analysis for customer feedback

2️⃣ Model Building

• Feature preprocessing
• Classification model training
• Model evaluation using:
• Accuracy
• F1-score
• Recall
• ROC Curve

3️⃣ Prediction Generation

Final predictions saved in outputs/predictions.csv Best trained model saved as best_model.pkl

---

📌 Important Note About Model File

• ⚠️ The trained model file best_model.pkl is not uploaded to GitHub because the file size is too large for GitHub’s upload limits.
• 👉 All outputs, graphs, metrics, and prediction files are included in the repository.
• If needed, the model can be recreated by running the ML notebook.

---

🖥 How to Run This Project Locally

1️ Clone the Repository

git clone https://github.com/aparna190417/Customer-Satisfaction-Prediction.git cd Customer-Satisfaction-Prediction

2 Install Required Libraries

pip install -r requirements.txt

3 Run the Streamlit Dashboard

cd app streamlit run app.py

Run Model Test Script

python check_model.py

---

📊 Dashboard Features

• Executive KPI Snapshot
• Ticket trends & satisfaction analysis
• Voice of Customer (WordCloud)
• Model performance visuals
• Live AI satisfaction prediction
• Download prediction results as CSV

---

📈 Power BI Dashboard

The Power BI dashboard converts raw data and ML outputs into business-friendly insights.

Dashboard Pages:

• Overview
• Tickets Analysis
• ML Performance
• Data Health
• Prediction
• Voice of Customer

📌 Note: The .pbix file must be opened using Power BI Desktop.

---

🛠 Technologies Used

• Python
• Pandas, NumPy
• Matplotlib, Seaborn
• Scikit-learn
• NLP (Text features)
• Streamlit
• Power BI

---

🚀 Key Insights

• Identified customers likely to be dissatisfied
• Highlighted critical ticket channels and priorities
• Improved visibility into data quality and ML performance
• Enabled proactive customer support strategies

💡 Business Impact

• Supports data-driven customer experience decisions
• Helps reduce customer churn
• Bridges the gap between Machine Learning and business users

👩‍💻 Author

Aparna Patel