Skip to content

Latest commit

 

History

History
143 lines (123 loc) · 4.37 KB

README.md

File metadata and controls

143 lines (123 loc) · 4.37 KB

Travel Insurance Prediction Web App

Brief Explanation

This Project has a goal to predict whether a customer of travel insurance company is willing to buy its' insurance product. Dataset was downloaded from kaggle. Please upvote the dataset contributor if you guys do not mind.

The Project itself has theese following features :

  1. Exploratory Data Analysis Overview
  2. Custom Model Creation
  3. Prediction

Technology Stacks :

  1. Heroku for Deployment
  2. streamlit for web page creation
  3. Data processing packages / libraries such as Pandas, Numpy, seaborn , and etc.
  4. Machine Learning Model : sklearn, xgboost

if you curious about the web app you can check here

Project Workflow

  1. Data Exploration -- data cleaning-visualization-encoding
  2. Model Training -- splitting dataset-cross validation-final model-save model
  3. Creating interface -- layouting-adding menus
  4. Deploy!

Project Structure

In order to create this project i create several files including jupyter notebook and python scripts.

  1. modelling.ipynb for exploring the dataset from data cleaning to modelling
  2. webapp.py python script for creating streamlit based web app
  3. poetry.lock for dependency management
  4. Dockerfile for containerizing app

Metrics

Since the dataset's target class is not well distributed we attempt to use ROC AUC Scoring. The Model is using

xgb_param  = {
            'objective':'binary:logistic',
            'max_depth': 6,
            'alpha': 6,
            'learning_rate': 0.01,
            'n_estimators':400
        }  

voting_classifier_params = 
{'estimators': [('lgb', LGBMClassifier()),
  ('rf', RandomForestClassifier()),
  ('gbc', GradientBoostingClassifier()),
  ('cat', <catboost.core.CatBoostClassifier at 0x2ae0594bdf0>)],
 'flatten_transform': True,
 'n_jobs': None,
 'verbose': False,
 'voting': 'soft',
 'weights': None,
 'lgb': LGBMClassifier(),
 'rf': RandomForestClassifier(),
 'gbc': GradientBoostingClassifier(),
 'cat': <catboost.core.CatBoostClassifier at 0x2ae0594bdf0>,
 'lgb__boosting_type': 'gbdt',
 'lgb__class_weight': None,
 'lgb__colsample_bytree': 1.0,
 'lgb__importance_type': 'split',
 'lgb__learning_rate': 0.1,
 'lgb__max_depth': -1,
 'lgb__min_child_samples': 20,
 'lgb__min_child_weight': 0.001,
 'lgb__min_split_gain': 0.0,
 'lgb__n_estimators': 100,
 'lgb__n_jobs': -1,
 'lgb__num_leaves': 31,
 'lgb__objective': None,
 'lgb__random_state': None,
 'lgb__reg_alpha': 0.0,
 'lgb__reg_lambda': 0.0,
 'lgb__silent': True,
 'lgb__subsample': 1.0,
 'lgb__subsample_for_bin': 200000,
 'lgb__subsample_freq': 0,
 'rf__bootstrap': True,
 'rf__ccp_alpha': 0.0,
 'rf__class_weight': None,
 'rf__criterion': 'gini',
 'rf__max_depth': None,
 'rf__max_features': 'auto',
 'rf__max_leaf_nodes': None,
 'rf__max_samples': None,
 'rf__min_impurity_decrease': 0.0,
 'rf__min_impurity_split': None,
 'rf__min_samples_leaf': 1,
 'rf__min_samples_split': 2,
 'rf__min_weight_fraction_leaf': 0.0,
 'rf__n_estimators': 100,
 'rf__n_jobs': None,
 'rf__oob_score': False,
 'rf__random_state': None,
 'rf__verbose': 0,
 'rf__warm_start': False,
 'gbc__ccp_alpha': 0.0,
 'gbc__criterion': 'friedman_mse',
 'gbc__init': None,
 'gbc__learning_rate': 0.1,
 'gbc__loss': 'deviance',
 'gbc__max_depth': 3,
 'gbc__max_features': None,
 'gbc__max_leaf_nodes': None,
 'gbc__min_impurity_decrease': 0.0,
 'gbc__min_impurity_split': None,
 'gbc__min_samples_leaf': 1,
 'gbc__min_samples_split': 2,
 'gbc__min_weight_fraction_leaf': 0.0,
 'gbc__n_estimators': 100,
 'gbc__n_iter_no_change': None,
 'gbc__presort': 'deprecated',
 'gbc__random_state': None,
 'gbc__subsample': 1.0,
 'gbc__tol': 0.0001,
 'gbc__validation_fraction': 0.1,
 'gbc__verbose': 0,
 'gbc__warm_start': False}

if you want to customize on your own feel free to download model in my this web app

Model Name Accuracy Score ROC AUC Score
XGBClassifier 0.849246 0.849135
VotingClassifier 0.853162 0.849246

Further Improvement

  1. Serve API With FastAPI
  2. Moving to Dash for more interact look
  3. Implementing Artifical Neural Networks

What if you want to contribute

just fork my repository and star it