Used a tabular dataset for black friday sales prediction collected from one store. Trained Neural Network with three different regression models- a) Decision Tree regressor, b) Random Forest Regressor, and c) Gradient Boosting Regressor. Performed preprocessing of data by cleaning the data, performed indivisual variable analysis and multi variate analysis. Trained the data using three regressors and evaluated the model using RMSE score for the three models.

Implemented transfer learning model for increasing model accuracy for object detection using improved activation function. This improved model is implemented using open set deep networks, where the open set classifier is subjected to openMax activation function. First model for the transfer learning approach was trained on CIFAR-10 dataset to obtain weights to be used on MNIST dataset. Training of the model was performed using tensorflow deep networks with depth of 4 convolution layers. This model resulted in reducing false positives given during object detection.

This project aims to measure how reviews and ratings of movies released in theaters relate to their sales at the box office. The analysis addresses how the general sentiments of the public towards specific theatrical releases relate to box office results could bring substantial benefits to the movie industry. Four different datasets were used for the analytical prediction from the data collected were curatted using: 1)The Movies Dataset 2)IMDB movies dataset 3)US Unemployment dataset 4)Bureau of Labor Statistics Unemployment rates. Data was preprocessed and cleeaned for all the given datasets in order to combine them together. The combined data was analyzed to perform bivariate analysis and for Modelling. Modelling was implemented using both Supervised (Regression and Classification) and Unsupervised (K-Mean clustering) models.

Used PostgreSQL to create and manage dataset for airline booking information of costumers in this project. Also used java based web appplication for the front end implementation, included 20 rows of data in the PostgreSQL to store and process.

Analyzed different services provided by AWS for implementing machine learning models Analysis using different AWS technologies (S3, EC2). Designed and analyzed classifier machine learning models like: Regression, Decision Tree, SVM, Gradient Boost.

This project originated from an idea aimed at real world problem solving using Computer Vision. There were not many machine learning applications targeting oral cancer. Vision for the project: to make an accessible application to generate alerts for users to consult and get tested for oral cancer. Performed tasks like data collection, labeling, data cleaning, preprocessing of the images for the model. Multiple approaches were tested for this problem such as: Transfer learning, CNN. From testing of different approaches for results based on parameters like accuracy, error rate VGG-16 architecture was successfully implemented. Results from the model is a binary classification (If the given image is cancerous or not).

This is a hardware based device which is used in detecting white spaces where the applications range from finding the number of words on a paper to detecting white spaces in chess boards. This was arduino based project, objective was to detect black & white space in between characters & symbols.