# Wafer Classification
This application trains ML models on [wafer](https://en.wikipedia.org/wiki/Wafer_(electronics)) sensor data and classifies if the wafers are Good or Bad. The dataset contains wafer names and 590 columns of different sensor values for each wafer. Datasets for training and prediction are provided separately as a set of csv files. A 'schema' file is used to verify the data format. Logging is performed frequently and log files are stored in the Prediction and Training specific log folders. The code for this application can be found [__here__](https://gitfront.io/r/mohitagr18/e602ff783eb69e690d729a6ffecc3ee0da88e51e/wafer-detection/).

__Application Access:__
The application is deployed on [HEROKU](https://www.heroku.com/) and can be accessed via [POSTMAN](https://www.postman.com/)

- Prediction End Point: https://fathomless-brook-01494.herokuapp.com/predict
- Request Type: POST
- Header: {'Content-Type':'application/json'}
- Body: {"filepath":"Prediction_Batch_files"}

![image](https://user-images.githubusercontent.com/8663281/137374000-14c730bc-766d-46dc-a05b-d110258d5df8.png)


__Libraries Used:__
- App framework - flask, wsgiref
- DB operations - sqlite3
- Data processing and ML - numpy, pandas, matplotlib, sklearn, xgboost, kneed, pickle
- General operations - os, shutil, csv, json, re, datetime, time


![image](https://user-images.githubusercontent.com/8663281/137382144-607f116a-1b3a-42fd-ad69-6871ed34699c.png)