Detection-of-Neuroatypical-Disorder

Feature Exploration

Features	Adolescent Dataset	Adult Dataset	Child Dataset
Total number of records	104	704	292
Patients diagnoised with disorder	63	189	141
Patients not diagnoised with disorder	41	515	151
Percentage of patients diagnoised with disorder	60.58 %	26.85 %	48.29 %

Data Preprocessing

• Handle the missing values
• Split the data into features and target label
• Normalize the numerical variables using MinMax Scaler
• One-hot encoding for categorical variables
• Encoded the asd_classes column

Encoded ASD class dataset visualization

Child dataset:

Adult Dataset:

Adolescent Dataset:

Machine Learning Techniques

• Logistic Regression
• Decision Tree
• K-Nearest Neighbor
• Naive Bayes
• Sequential Model (CNN)

Accuracy, AUC score, cross validation and F-beta score comparison

Child Dataset:

Techniques	Accuracy	Precision	Sensitivity	F-beta score	F1 score
Logistic Regression	0.98783	0.966	0.95	0.95	0.95
Decision Tree	1	1	1	1	1
K-Nearest Neighbor	0.96	0.933	1	0.94594	0.966551
Naive Bayes	0.86	0.838	0.928	0.85526	0.8813
Sequential Model (CNN)	0.959	0.964	0.964	0.964	0.964

Adult Dataset:

Techniques	Accuracy	Precision	Sensitivity	F-beta score	F1 score
Logistic Regression	0.9836	0.956	1	0.9641	0.97727
Decision Tree	1	1	1	1	1
K-Nearest Neighbor	0.959	0.9318	0.9534	0.9360	0.94252
Naive Bayes	0.8934	0.8260	0.8837	0.837	0.8539
Sequential Model (CNN)	0.97540	0.95454	0.9767	0.9589	0.9655

Adolescent Dataset:

Techniques	Accuracy	Precision	Sensitivity	F-beta score	F1 score
Logistic Regression	0.9	0.92307	0.92307	0.92307	0.92307
Decision Tree	1	1	1	1	1
K-Nearest Neighbor	0.75	0.7222	1	0.7647	0.8387
Naive Bayes	0.75	0.7222	1	0.76470	0.8387
Sequential Model (CNN)	0.699	0.8181	0.6923	0.78947	0.75001

Confusion Matrix and Learning curves

Child Dataset:

Adult Dataset:

Adolescent Dataset:

License

The detection project is free and open-source software licensed under the Apache-2.0 license.