Machine_learning_projects

Overview:-

• This repository contains projects on ML Classification and Regression

• Classification:- classification is the problem of identifying to which of a set of categories (sub-populations) a new observation belongs, on the basis of a training set of data containing observations (or instances) whose category membership is known.

• Ex. Prediction of heart disease( Yes or No)

• Regression:- In statistical modeling, regression analysis is a set of statistical processes for estimating the relationships between a dependent variable (often called the 'outcome variable') and one or more independent variables (often called 'predictors', 'covariates', or 'features').

• Ex. Prediction of number (price prediction)

Repository Guide:-

Here is list for every project in this repository and link to introduction of project

---

Classification:-

• Telco Customers Churn Prediction
• Heart Disease Prediction
• Titanic Survival Prediction

Regression:-

• Bulldozers Price Prediction
• House Price Prediction
• Flight Fare Prediction (Web-App)

Tools And Workflow:-

• Collect Data
• Exploratory Data Analysis:- (with visualization)

• Pandas, Matplotlib, Seaborn

• Feature Engineering:-

• Handle missing values and Categorical Features - sklearn
• Outlier's (depending on which model to use)
• Feature Scaling (depending on model)

• Split data
• Feature selection:-

• Correlation matrix, VariationThreshold

• Building model

• Selecting models
• Train and Evaluate model
• If more than one depending on initial accuracy take for Hyperparameter Tuning - RandomSearchCV, GridSearchCV

• Predict on test set
• Build web-app Flask, HTML5, CSS3 or Streamlit
• Deploy on cloud

Project Introduction's:-(for detail discription visit respective project)

Classification:-

---

Telco Churn Predictor

• This project is to predict if customer will leave telco or not, using machine learning
• In this we used XGBoost classifier to predict it and got 78.64 accuracy

• by using gamma=0, learning_rate=0.1, max_depth=4, reg_lambda=10, scale_pos_weight=2, subsample=0.9, colsample_bytree=0.5, which we got by hyperparameter tunning XGBClassifier

• Get source code Visit

Heart Disease Predictor

• On basis of person's health records predict if person have heart disease or not
• In this we used 4 different models to see initial accuracy (ie. SVM, RandomForestClassifier, Adaboost, KNN)

• Out of this we we picked top two for hypertunning( RandomForest and Adaboost) and Adaboost got 91% accuracy_score

• Get source code Visit

Titanic Survial Predictor

• This project is to predict the passangers in titanic will survive or not on the basis of given data.
• In this notebook we've used Random Forest,KNN and GradientBoost classifier and got better accuracy on these top 2 models i.e.GradientBoostClassifier and RandomForestlassifier.
• After Hypertunning got an accuracy 78.73% for Gradient Boosting.

• Hypertunnig parameters for GBC learning_rate= 0.01, max_depth= 4, max_features= 16, min_samples_leaf= 1, min_samples_split=20, n_estimators= 500.

• Get source code Visit

Regression:-

---

Bulldozer Price Predictor

• Prediction of sale price of bulldozer on basis of its specifications like year made, productSize and 50 more features
• Here we used RandomForestRegressor and got pretty good score of 85% r^2
• so after hypertuning we got best_params n_estimators=90, min_samples_leaf=1, min_samples_split=14,max_features=0.5, n_jobs=-1, max_samples=None
• with this we got 95% and 88% r^2 on train and test set respectively
• Get source code Visit

House Price Predictor

• Prediction of House price on basis of features of house
• Here we used Lasso, randomforest and XGBoost for initial r^2 score
• Then we hypertuined XGB cause it had bit more r^2 and got gamma= 0, learning_rate= 0.05, max_depth =6, reg_lambda= 0, subsample=0.9, colsample_bytree=0.5
• with this params I got 98% and 85% r^2 on train and validation set respectively
• Get source code Visit

Flight Fare Prediction

• Predicting fare of flight based on given information
• Trained Lasso, Gradient Bosst, Random Forest on full data
• Random forest had best 80 r2 so took this for hyperparameter tuning got 83 r2
• Using Flask with help of html5-css3 built web-app and Deployed on Heroku
• To view demo of web app visit

---