PCOS_Classification_Project

Based on levels of certain vitamins, hormone levels and physical activities, predict if a woman may have PCOS

Objectives:

1. Analyse the PCOS dataset [https://www.kaggle.com/datasets/prasoonkottarathil/polycystic-ovary-syndrome-pcos] ✔️
2. Perform Binary Classification task using following ML Algorithms: Logistic Regression, SVM, Decision Trees, Random Forest, Naive Bayes and K-Nearest Neighbours ✔️
3. Build Neural Network Model to perform Classification task. ✔️
4. Using the trained Neural Model deploy it using Web Interface. ⏳

1. Analysis [PCOS_Dataset_Analysis.ipynb]

1. PCOS Women have a majority of Cycle (R/I) based around 4 days, while non-PCOS Women have a majority of Cycle (R/I) based around 2 days

2. Cycle Lengths among PCOS women have been recorded to be around 2-8 days.

3. The age of PCOS Women range from 20-45 yrs, with majority cases between 25-30 yrs.

4. Non-PCOS Women have follicle lengths centered around 5 units, while PCOS Women have their follicle lengths centered around 10 units of measurement.

5. PCOS Women have a FSH/LH ratio between -5 to 35, while non-PCOS women have between 0-59.

6. PCOS Women have Vit D3 (ng/mL) levels between -100 to 90 as compared to non-PCOS Women whose Vit D3 (ng/mL) levels are centered around 10 to 70.

7. PCOS Women have a PRG(ng/mL) level between 0 to 1, while non-PCOS women have PRG levels between -0.5 to 1.5

8. PCOS Women also experience weight gain, hair-growth, skin-darkening, hair-loss, pimples and craving for fast-foods after developing PCOS.

2. Classification using ML Algorithms [PCOS_Classification.ipynb]

1. Out of the 6 algorithms used, Naive Bayes performed the best with the highest accuracy (83.7%) and highest precision(94.6%) in comparison to other models.

   Models	Accuracy	Precision	Recall
   Logistic Regression	0.7703	0.5178	0.8787
   Support Vector Machines	0.6666	0.3928	0.6666
   Decision Trees	0.7777	0.6250	0.7954
   Random Forest	0.8222	0.6071	0.9444
   Naive Bayes	0.8370	0.9464	0.7361
   K-Nearest Neighbours	0.6000	0.1964	0.5500

2. After feature selection using Wrapper method (Backward Elimination), 18 best features were selected for training with Naive Bayes model which led to the following results:

   Metric	Score
   Accuracy	0.851852
   Precision	0.892857
   Recall	0.78125

3. Neural Network Performance [PCOS_CLassification_NeuralNetwork.ipynb]

Constructed a Neural Network model with PyTorch to classify features into PCOS and Non-PCOS categories. The following is the architecture of the Neural Model:

The Test Results obtained after the result are as follows:

Class (Yes: 1 / No: 0)	Precision	Recall	F1-Score	Support
0	0.94	0.96	0.95	106
1	0.92	0.88	0.90	56
Accuracy	-	-	0.93	162

4. Web Integration

(work in progress)