Hybrid Cricket Intelligence Model (CIM)

A Streamlit-based AI/ML toolkit for IPL team strategy — providing Best XI selection (Genetic Algorithm), player performance predictions (SVM), opponent matchup insights (Apriori), and bowler-batsman dismissal probability (XGBoost).

---

Quick Overview

• Language: Python
• UI: Streamlit
• Main features:
  • GA Team Selection: generate a Best XI for a franchise and venue
  • SVM Player Performance: predict player performance categories
  • Apriori Opponent Analysis: find historical matchup patterns
  • XGBoost Matchup Prediction: compute bowler-vs-batsman dismissal probabilities

This repository contains the code, data references, images, and pre-trained model artifacts required to run a local interactive dashboard.

---

Repository Layout (important files)

• CIM/
  • models/ - The main Streamlit app and model scripts
    • app.py - Main Streamlit application (tabs: GA, SVM, Apriori, XGBoost)
    • ga_team_selector.py - Genetic algorithm implementation (Best XI)
    • ap_opponent_analysis.py - Apriori-based opponent analysis helper
    • svm_player_performance.py - SVM model code and helper functions
    • train_xgb_delivery.py - XGBoost training utilities (if present)
    • player_roles_cleaned.csv - cleaned roles reference
  • data/ - CSV data used by the app
    • player_stats_venue.csv
    • player_roles.csv
    • ball_by_ball_stats_ap.csv
    • match_stats.csv and others used for analyses
  • pictures/
    • logos/ - team logo images (referenced by app.py)
    • ipl_stadium_bg.jpg - background image
    • ipl_logo.png - app header logo
  • static/ - static assets used by the UI

Top-level CSV(s):

• player_stats_venue.csv (also present at repository root in some copies)

---

Prerequisites

• Python 3.8+ recommended
• Install required packages (example):

pip install streamlit pandas numpy scikit-learn xgboost joblib matplotlib seaborn

(If you have a requirements.txt file, prefer pip install -r requirements.txt.)

---

Configuration

app.py contains a number of hard-coded paths near the top. Default values used in the app:

STATS_FILE = "D:/AI ML Cricket Project CIM model/CIM/data/player_stats_venue.csv"
ROLES_FILE = "D:/AI ML Cricket Project CIM model/CIM/data/player_roles.csv"
BALL_BY_BALL_FILE = "D:/AI ML Cricket Project CIM model/CIM/data/ball_by_ball_stats_ap.csv"
BG_PATH = "D:/AI ML Cricket Project CIM model/CIM/pictures/ipl_stadium_bg.jpg"
LOGO_PATH = "D:/AI ML Cricket Project CIM model/CIM/pictures/ipl_logo.png"
LOGOS_PATH = "D:/AI ML Cricket Project CIM model/CIM/pictures/logos"

# Model artifact names (expected in models/ working dir)
SVM_BATTER = "svm_batter_model.pkl"
SVM_BOWLER = "svm_bowler_model.pkl"
SVM_FEAT = "svm_player_performance_features.pkl"
XGB_MODEL = "xgb_delivery_model.pkl"
XGB_ENC = "xgb_delivery_label_encoders.pkl"
XGB_FEAT = "xgb_delivery_features.pkl"

If your working directories differ, update these constants at the top of app.py before running.

---

Running the App

1. Open a terminal and change to the CIM/models directory:

cd "D:\AI ML Cricket Project CIM model\CIM\models"

2. Run Streamlit:

python -m streamlit run app.py

3. The app will open in your browser (or use the local URL shown in the terminal). Navigate between tabs to run:

• Tab 1: GA Team Selection — select Franchise + Venue, click "Generate Best XI"
• Tab 2: SVM Player Performance — compute player performance predictions for the Best XI
• Tab 3: Apriori Opponent Analysis — select opponent and XI to run Apriori
• Tab 4: XGBoost Matchup Prediction — compute bowler-batsman dismissal probabilities (relies on Best XI + Opponent & Venue)

Notes:

• The Apriori opponent selection is persisted to session state and used by the XGBoost tab to display the opponent badge.
• Some tabs require that the GA (Tab 1) has been run first (Best XI available in session).

---

Data & Model Files

• Data CSVs must be present at the paths configured in app.py. If data files are large, place them in the CIM/data/ folder and update the constants.
• Pre-trained model artifacts (SVM/XGBoost) must be present in CIM/models/ or update the path variables. If you need to re-train models, use the training scripts in CIM/models/ if available.

---

UI & Styling

• app.py injects a custom CSS block via inject_css() for a premium dashboard look. If you need to revert to simple styling, remove or edit that function.
• Team logos are loaded from LOGOS_PATH. Team-to-logo mapping is defined in app.py via the TEAM_LOGO_MAP dictionary.
• The app uses Streamlit session state keys (e.g. best_xi, input_team_display, opponent_team, xgb_matchups) to pass information across tabs.

---

Troubleshooting

• Common errors:

  • File not found: verify the CSV and image paths in app.py.
  • Model loading error: ensure .pkl files exist and are compatible with your Python package versions.
  • Streamlit caching issues: try restarting the app and clearing Streamlit cache (streamlit cache clear for older versions or delete .streamlit cache folder).

• If the UI shows raw </div> text or other stray HTML: this was caused previously by manually writing closing tags via st.markdown("</div>", unsafe_allow_html=True). The app has been updated to avoid rendering stray closing tags; if you see these, ensure you are running the latest app.py.

---

Developer Notes

• To add a new team logo: copy an image into CIM/pictures/logos/ and add the filename to the TEAM_LOGO_MAP dictionary in app.py.
• Session keys of interest:
  • st.session_state['best_xi'] — pandas DataFrame of best XI
  • st.session_state['input_team_display'] — the franchise selected in GA
  • st.session_state['opponent_team'] — selected opponent from Apriori tab (used by XGBoost)
  • st.session_state['xgb_matchups'] — latest XGBoost matchup DataFrame

---

Contribution

• Fork and create feature branches for changes.
• Keep UI and CSS changes isolated, and avoid embedding raw closing HTML tags via st.markdown (Streamlit sometimes renders them as text).

---

Contact

Maintainer: Santhoji V

If you'd like help running the app or customizing visuals, open an issue or contact the maintainer directly.

---

Generated README — adapt paths and instructions to your local environment as needed.