🏏 IPL First Innings Score Prediction

This is one of my favorite end-to-end machine learning projects where I worked on predicting the first-innings score in an IPL match. I personally handled everything – from data collection and preprocessing, to model training, evaluation, and finally deploying the solution as a Flask web application.

🔗 Demo Video: Watch here 📷 Web App Screenshot: 📓 Jupyter Notebook: View Here

---

📌 Objective

My goal was to build a model that predicts the final first-innings score of an IPL match using real-time match data such as:

• Batting and bowling teams
• Current runs, wickets, overs
• Runs and wickets in the last 5 overs

---

🗂️ Project Workflow

This project follows a complete ML pipeline:

1. Data Collection:

   • Collected ball-by-ball IPL data (2008–2017) from Kaggle.

2. Data Preprocessing:

   • Cleaned dataset by removing unnecessary columns like mid, venue, batsman.
   • Filtered consistent teams and dropped overs less than 5 for meaningful stats.
   • Performed categorical encoding using get_dummies().
   • Split data into training (till 2016) and testing (2017) sets.

3. Model Training:

   • Trained a Linear Regression model using Scikit-learn.
   • Achieved an R² score of ~82%, which accurately captures scoring trends.

4. Model Saving:

   • Saved the trained model with Python’s pickle as lr-model.pkl.

5. Flask Deployment:

   • Developed a user-friendly Flask web app with HTML frontend.
   • Users can select teams, overs, runs, wickets, etc., and instantly get predicted scores.

---

⚙️ Technologies Used

• Python, Pandas, NumPy – Data processing
• Matplotlib – Visualization
• Scikit-learn – Machine Learning (Linear Regression)
• Flask – Web deployment
• HTML/CSS – Frontend UI

---

🚀 Sample Prediction

Example Input:

• Batting team: Mumbai Indians
• Bowling team: CSK
• Overs completed: 12.3
• Runs: 105
• Wickets: 3
• Runs in last 5 overs: 45
• Wickets in last 5 overs: 2

Predicted Final Score: 178 Runs

---

🧪 Model Performance

Metric	Result
Algorithm	Linear Regression
Train/Test Split	80/20
Evaluation Metric	R² Score
Accuracy	~82%

---

👨‍💻 My Role

I designed, implemented, and deployed the entire project myself — from data engineering, feature engineering, and model training, to building the Flask web app and deploying it locally for demonstration.

---