Machine Learning Code Using Python

This project contains various Machine Learning algorithms implemented using python 3.0 compiled on jupyter notebook.

Overview

Each folder containg .ipynb file. Some of the dataset used were imported from library and some are uploaded in the folders as .csv file.

Machine Learning Algorithms:

1) Linear Regression

Project description

E Commerce Data

• Worked on E Commerce data to predict Yearly amount spent by Customer on app or on website
• Decided which factor affects the amount of money spent by customer, few of the factors are: Avg. Session Length, Time on App, Time on Website & Length of Membership
• Data File: Ecommerce Customers.csv

Housing Data

• Worked on USA Housing data to predict price of house using various factors such as: Avg. Area Income, Avg. Area House Age, Avg. Area Number of Rooms, Avg. Area Number of Bedrooms & Area Population
• Data File: USA_Housing.csv

2) Logistic Regression

Titanic Data

• Worked on Titanic data frame to predict survival or non survival of people on titanic
• Used mean to fill missing values
• Created dummy variables for categorical variables
• Data File: titanic_train.csv, titanic_test.csv

Advertising Data

• Worked on Advertising data to predict whether customer clicked on advertisement on not based on data available about various features
• Data File: advertising.csv

3) Decision Trees & Random Forest

LendingClub.com data

• Worked on Lending club data to predict if a borrower pays the money back to investor or not, using decision tree and random forest classifier
• Data File: loan_data.csv

Kyphosis Data

• Predicted whether a person has Kyphosis or not using decision tree and random forest classifier based on data provided
• Data File: kyphosis.csv

4) K Means Clustering

University Data

• Used K mean clustering to clusters universities as Private or Public based on various features provided
• Data File: College_Data.csv

Blob Data

• Used K mean clustering to clusters Blobs into 4 clusters

5) K nearest Neighbor Classifier

Classified Data

• Worked on classified data set from kaggle to do K Nearest Neighbor Classification
• Used StandardScaler() to scale the data
• Data File: Classified

6) Support Vector Machine

Breast Cancer Data

• Worked on Breast cancer data to predict whether a cell is cancer cell or not using Support Vector Machine (SVM) and GridSearchCV to predict parameters for C and gamma best for our model

Iris Flower Data Set

• Worked on Iris data set to predict various species of flower using Support Vector Machine (SVM) and GridSearchCV to predict parameters for C and gamma best for our model

7) Movie Recommender System

Movie Data

• Worked on Movie rating data from various users to find highly correlated movies to each other (according to ratings provided by different users)
• Data File: Movie_Id_Titles, u.data

8) Natural Language Processing

Yelp Data Set from Kaggle

• Worked on Yelp data set to predict stars provided by reviewer using length of the text
• Created a function to remove the punctuation from the review and used stopwords() to remove stop words such as 'i', 'me', 'my', 'myself' etc.
• Used CountVectorizer to convert reviews in vectors of different words
• Using TfidfTransformer created inverse document frequency
• After that created prediction model using Multinomial Naive Bayes
• Data File: yelp.csv

Ham vs Spam

• Same steps were used as Yelp project but data was used from ('smsspamcollection/SMSSpamCollection')

9) Tensor Flow: MNIST Data set

MNIST Data

• Worked on MNIST Data set and predicted the numerical value from images using TensorFlow library in Python
• Data Set: MNIST

10) Ridge & Lasso Regression

• Worked on Data set and chose best parameter for Ridge and Lasso Regression using cross validation
• Data Set: test.csv, train.csv

11) Sentiment Analysis On Amazon’s Product

• The project is based on providing any user, using Amazon (any Ecommerce site), identifying and decision making within initial 10 seconds on whether to buy a product or not based on user reviews given on the website.

• Mainly, using Natural Language Processing and Sentiment Analysis in identifying whether a review by a user is positive or not. Algorithm helps any user who goes to any product page, extract all the reviews given and perform sentiment analysis on all the reviews and give a numeric prediction like out of say 10000 reviews 3000 are positive, 1500 are neutral and 5500 are negative.

• This will help the user in identifying whether to go for that product within 5 to 10 seconds instead of spending multiple minutes reading description, ratings, reviews etc.

Prerequisites

• Python 3.0
• Jupyter Notebook
• Libraries: numpy, sklearn, pandas, seaborn, matplotlib, tensorflow