Project : Churnify

• This project is in development
• This project is a part of my graduation - thesis project

---

• This project is being developed using flask-restful, pymongo, boto3, scikit-learn, tensorflow and keras
• Project uses a NoQSL MongoDB to store users and models details
• This project runs 6 different classification algorithm from scikit-learn and neural network with different parameters from keras

USED SERVICES

• Heroku to publish python-flask server app
• MongoDB cloud for nosql database
• AWS S3 for file storage
• Keras for neural network construction
• Tensorflow as backend to Keras

METHODS

It has following methods that can be requested:

• MainPage ()

• Test ()

• Register (email, username, password)

• Login (username, password)

• GetUserPlan (username, password)

• UpdateUserPlan (username, password, usertype)

• ColumnsInfos (username, password, columns, dataset)

• Train (username, password, modelname, dataset, columns, target, categoricalcolumns, numericalcolumns, modelType = None, {classifier specific variables})

• Predict (username, password, modelname, predictset)

• ModelList(username, password)

• CheckTrainStatus (username, password)

• RemoveModel (username, password, modelname)

GMS Module Details

Also there is GMS(Generative Model Selector) class working in background to find best algorithm and parameters for given dataset

GMS has 8 different classification algorithm to try with different parameters:

• Logistic Regression
• KNN
• Naive Bayes
• Kernel SVM
• Decision Tree
• Random Forest
• Neural Network
• XGBoost

Live version of this code is in:

https://churn-analysis-api.herokuapp.com/

You can send requests to following URLs:

GET https://churn-analysis-api.herokuapp.com/

GET https://churn-analysis-api.herokuapp.com/test

POST https://churn-analysis-api.herokuapp.com/register

POST https://churn-analysis-api.herokuapp.com/login

POST https://churn-analysis-api.herokuapp.com/getUserPlan

POST https://churn-analysis-api.herokuapp.com/updateUserPlan

POST https://churn-analysis-api.herokuapp.com/columnsInfos

POST https://churn-analysis-api.herokuapp.com/train

POST https://churn-analysis-api.herokuapp.com/predict

POST https://churn-analysis-api.herokuapp.com/modelList

POST https://churn-analysis-api.herokuapp.com/checkStatus

POST https://churn-analysis-api.herokuapp.com/removeModel

TODOs:

• Using Cross Validation metrics to find best model and eliminate overfitted models
• Grid Search to find best parameters
• Optimization
• Encapsulation