🌟 Autism Spectrum Disorder (ASD) Detection using Machine Learning

🎯 AIM

To develop a machine learning model that predicts the likelihood of Autism Spectrum Disorder (ASD) based on behavioral and demographic features.

🌊 DATASET LINK

Autism Screening Data

📚 KAGGLE NOTEBOOK

Autism Detection Kaggle Notebook

Kaggle Notebook

⚙️ TECH STACK

Category	Technologies
Languages	Python
Libraries/Frameworks	Pandas, NumPy, Scikit-learn,
Tools	Jupyter Notebook, VS Code

🖍 DESCRIPTION

What is the requirement of the project?

The rise in Autism cases necessitates early detection.
Traditional diagnostic methods are time-consuming and expensive.
Machine learning can provide quick, accurate predictions to aid early intervention.

How is it beneficial and used?

Helps doctors and researchers identify ASD tendencies early.
Reduces the time taken for ASD screening.
Provides a scalable and cost-effective approach.

How did you start approaching this project? (Initial thoughts and planning)

Collected and preprocessed the dataset.
Explored different ML models for classification.
Evaluated models based on accuracy and efficiency.

🔍 PROJECT EXPLANATION

🧩 DATASET OVERVIEW & FEATURE DETAILS

The dataset consists of 800 rows and 22 columns, containing information related to autism spectrum disorder (ASD) detection based on various parameters.

Feature Name	Description	Datatype
`ID`	Unique identifier for each record	`int64`
`A1_Score` - `A10_Score`	Responses to 10 screening questions (0 or 1)	`int64`
`age`	Age of the individual	`float64`
`gender`	Gender (`m` for male, `f` for female)	`object`
`ethnicity`	Ethnic background	`object`
`jaundice`	Whether the individual had jaundice at birth (`yes/no`)	`object`
`austim`	Family history of autism (`yes/no`)	`object`
`contry_of_res`	Country of residence	`object`
`used_app_before`	Whether the individual used a screening app before (`yes/no`)	`object`
`result`	Score calculated based on the screening test	`float64`
`age_desc`	Age description (e.g., "18 and more")	`object`
`relation`	Relation of the person filling out the form	`object`
`Class/ASD`	ASD diagnosis label (`1` for ASD, `0` for non-ASD)	`int64`

This dataset provides essential features for training a model to detect ASD based on questionnaire responses and demographic information.

🛠 PROJECT WORKFLOW

Project workflow

  graph LR
    A[Start] --> B[Data Preprocessing];
    B --> C[Feature Engineering];
    C --> D[Model Training];
    D --> E[Model Evaluation];
    E --> F[Deployment];

Step 1Step 2Step 3Step 4Step 5

Collected dataset and performed exploratory data analysis.

Preprocessed data (handling missing values, encoding categorical data).

Feature selection and engineering.

Trained multiple classification models (Decision Tree, Random Forest, XGBoost).

Evaluated models using accuracy, precision, recall, and F1-score.

🖥️ CODE EXPLANATION

Section 1: Data PreprocessingSection 2: Model Training

Loaded dataset and handled missing values.

Implemented Logistic Regression and Neural Networks for classification.

⚖️ PROJECT TRADE-OFFS AND SOLUTIONS

Trade Off 1Trade Off 2

Accuracy vs. Model Interpretability: Used a Random Forest model instead of a deep neural network for better interpretability.

Speed vs. Accuracy: Chose Logistic Regression for quick predictions in real-time applications.

🖼 SCREENSHOTS

Visualizations and EDA of different features

Age Distribution

Model performance graphs

Confusion Matrix

Features Correlation

Feature Correlation Heatmap

📉 MODELS USED AND THEIR EVALUATION METRICS

Model	Accuracy	Precision	Recall	F1-score
Decision Tree	73%	0.71	0.73	0.72
Random Forest	82%	0.82	0.82	0.82
XGBoost	81%	0.81	0.81	081

✅ CONCLUSION

🔑 KEY LEARNINGS

Insights gained from the data

Behavioral screening scores are the strongest predictors of ASD.
Family history and neonatal jaundice also show correlations with ASD diagnosis.

Improvements in understanding machine learning concepts

Feature selection and engineering play a crucial role in medical predictions.
Trade-offs between accuracy, interpretability, and computational efficiency need to be balanced.

🌍 USE CASES

Early ASD ScreeningAssistive Diagnostic Tool

Helps parents and doctors identify ASD tendencies at an early stage.

Can support psychologists in preliminary ASD assessments before clinical diagnosis.