Vitiligo Predictor

Predicts possible Vitiligo cases using the following features:

• age (in years),
• sex (M for male, F for female),
• history (Yes if any relative has white patches in skin, otherwise No)
• skin type (either I,II, III, IV or V)
• reading : Light reflectance of epidermis (in lux)

Trained Models

Model	Accuracy	Precision	Recall	F1-Score
AdaBoostClassifier(n_estimators=100)*	98.90%	97.52%	96.91%	97.21%
MLPClassifier(hidden_layer_sizes=(1000,100), max_iter=1000)	98.04%	96.20%	93.83%	95.00%
KNeighborsClassifier(n_neighbors=12)	96.94%	96.60%	87.65%	91.91%

* Used for prediction

Install Instructions

1. Install Pipenv if not installed yet

$ pip3 install pipenv 

2. Clone this repo

$ git clone https://github.com/gfabia/vitiligopredictor.git

3. Install requirements

$ cd vitiligopredictor
$ pipenv shell
$ pip install -r requirements.txt

Usage

Input Format

Input is passed via the command-line and must be in JSON format. Required parameters are: id (string), age (integer), sex (string), history (string), skin_type (string) and reading (integer).

Example:

[{"id": "R89", "age": 19, "sex": "M", "history": "No", "skin_type": "IV", "reading": 75}]

Multiple records are allowed. For example:

[{"id": "R89", "age": 19, "sex": "M", "patches": "No", "history": "No", "skin_type": "IV", "reading_location": "Left Toe and Leg #5", "reading": 75}, {"id": "R90", "age": 85, "sex": "F", "patches": "Yes", "history": "Yes", "skin_type": "IV", "reading_location": "Left Toe and Leg #5", "reading": 76}, {"id": "R91", "age": 90, "sex": "F", "patches": "Yes", "history": "Yes", "skin_type": "IV", "reading_location": "Left Toe and Leg #5", "reading": 76}]

Output Format

Output is printed on STDOUT, also in JSON format. If input data is valid then a prediction key would be added to each record. No means Normal and Yes means possible Vitiligo case.

For example:

{"success": true, "data": [{"id": "R89", "age": 19, "sex": "M", "history": "No", "skin_type": "IV", "reading": 75, "prediction": "No"}]}

Sample Run

$ ./vitiligopredictor '[{"id": "R89", "age": 19, "sex": "M", "history": "No", "skin_type": "IV", "reading": 75}]'
{"success": true, "data": [{"id": "R89", "age": 19, "sex": "M", "history": "No", "skin_type": "IV", "reading": 75, "prediction": "No"}]}

$ ./vitiligopredictor '[{"id": "R89", "age": 19, "sex": "M", "patches": "No", "history": "No", "skin_type": "IV", "reading_location": "Left Toe and Leg #5", "reading": 75}, {"id": "R90", "age": 85, "sex": "F", "patches": "Yes", "history": "Yes", "skin_type": "IV", "reading_location": "Left Toe and Leg #5", "reading": 76}, {"id": "R91", "age": 90, "sex": "F", "patches": "Yes", "history": "Yes", "skin_type": "IV", "reading_location": "Left Toe and Leg #5", "reading": 76}]'
{"success": true, "data": [{"id": "R89", "age": 19, "sex": "M", "patches": "No", "history": "No", "skin_type": "IV", "reading_location": "Left Toe and Leg #5", "reading": 75, "prediction": "No"}, {"id": "R90", "age": 85, "sex": "F", "patches": "Yes", "history": "Yes", "skin_type": "IV", "reading_location": "Left Toe and Leg #5", "reading": 76, "prediction": "No"}, {"id": "R91", "age": 90, "sex": "F", "patches": "Yes", "history": "Yes", "skin_type": "IV", "reading_location": "Left Toe and Leg #5", "reading": 76, "prediction": "Yes"}]}

$ ./vitiligopredictor '[{"id": "R89", "age": 19, "sex": "M", "history": "No", "skin_type": "VI"}]'
{"success": false, "error": "Column 'reading' is required."}

$ ./vitiligopredictor '[{"id": "R89", "age": 19, "sex": "M", "history": "Of course", "skin_type": "VI", "reading": 75}]'
{"success": false, "error": "In 'R89', got an invalid datum 'Of course'."}