Autism-Spectrum-Detection

A COMPARITIVE STUDY OF NATURE-INSPIRED ALGORITHMS FOR AUSTISM SPECTRUM DISORDER(ASD) DETECTION

Problem Statement

Perform a comparative study of five nature-inspired optimization algorithms for detecting autism in young adults.

Dataset

The Autism Spectrum Disorder (ASD) screening dataset is used for detecting autism in young adults. This dataset is obtained from UCI machine learning repository. The following link can be used to download this dataset –

http://archive.ics.uci.edu/ml/machine-learning-databases/00426/

Implementation Details

In this comparative study, we plan to implement five different nature-inspired optimization algorithms and compare their results with backpropagation and particle swarm optimization (PSO). The five algorithms that we will be implementing are as follows:

• Cuckoo Optimization
• Cat Swarm Optimization
• Grey Wolf Optimization
• Firefly Algorithm
• Bat Algorithm

A python library SwarmPackagePy will be used for implementing these optimization approaches. We intend to design a feed-forward neural network model and find out the best weights of the neural net model using these different approaches.

Expected Outcome

A comparative study of accuracies, errors and other metrics which will give an overview of the best nature-inspired optimization algorithm for detecting ASD.

Files Description

• Bat Algorithm : Contains implementation of Bat Algorithm, a snapshot of results and a snapshot of behaviour of the minimization function.

• Cat Swarm Optimization : Contains implementation of Cat Swarm Optimization, a snapshot of results and a snapshot of behaviour of the minimization function.

• Cuckoo Search Optimization : Contains implementation of Cuckoo Search Optimizatio, a snapshot of results and a snapshot of behaviour of the minimization function.

• Data Preprocessing : Contains code for preprocessing the raw data.

• Datasets : Contains all the raw datasets and the final cleaned dataset under the name of "Final_Dataset.csv".

• Firefly Algorithm : Contains implementation of Firefly Algorithm, a snapshot of results and a snapshot of behaviour of the minimization function.

• Gray Wolf Optimization : Contains implementation of Gray Wolf Optimization, a snapshot of results and a snapshot of behaviour of the minimization function.

• Neural Network Model : Contains implementation of Backpropagation Algorithm and snapshots of results, epochs vs loss and epochs vs accuracy graphs

• Particle Swarm Optimization : Contains implementation of Particle Swarm Optimization and a snapshot of the result.

Final Report

The final report of the entire analysis and implementation of the project can be found here