Overview §
A machine learning model for early autism screening using questionnaire-based data, incorporating demographic features and clinical indicators. Logistic Regression was identified as the most effective model, achieving ~86.9% accuracy with high recall—ideal for preliminary screening, especially in resource-limited settings.
Techniques Used §
- 🗂️ Data preprocessing & feature engineering: Cleaned and structured Kaggle-based questionnaire data including age, gender, medical history, and screening responses.
- 🤖 Model evaluation: Compared multiple classifiers—Logistic Regression, SVM, Random Forest, Naive Bayes—with cross-validation and hyperparameter tuning.
- ✅ Focus on recall: Minimized false negatives to improve reliability for ASD detection.
- 💻 Deployment-ready: Prepared for a user-friendly interface using Streamlit for future adoption.
Results §
- Achieved ~86.9% accuracy using Logistic Regression.
- High recall ensures most autism cases are correctly identified, minimizing false negatives.
- Logistic Regression outperformed SVM, Random Forest, and Naive Bayes in preliminary screening.
- Demonstrated reliability on questionnaire-based datasets, making it suitable for early detection in resource-limited settings.
- Model is deployment-ready for integration with a user-friendly interface (e.g., Streamlit).