Resume-Job-Description-Matching

The purpose of this project was to defeat the current Application Tracking System used by most of the organization to filter out resumes. In order to achieve this goal I had to come up with a universal score which can help the applicant understand the current status of the match. The following steps were undertaken for this project 1) Job Descriptions were collected from Glass Door Web Site using Selenium as other scrappers failed 2) PDF resume parsing using PDF Miner 3) Creating a vector representation of each Job Description - Used word2Vec to create the vector in 300-dimensional vector space with each document represented as a list of word vectors 4) Given each word its required weights to counter few Job Description specific words to be dealt with - Used TFIDF score to get the word weights. 5) Important skill related words were given higher weights and overall mean of each Job description was obtained using the product for word vector and its TFIDF scores 6) Cosine Similarity was used get the similarities of the Job Description and the Resume 7) Various Natural Language Processing Techniques were identified to suggest on the improvements in the resume that could help increase the match score

Pokemon-GO-Analytics

Pokemon Go! became a very famous augmented reality (AR) game in 2016 summer. In this project, we wanted to understand the success of the mobile app game. Specifically, the purposes of this project were (1) To do web scraping using BeautifulSoup, (2) To construct a Pandas dataframe (3) To explore/visualize the numeric data using seaborn (4) To use sklearn to build machine learning models to predict the app’s review counts using Linear, Ridge, Lasso, Elastic Net regression optimizing vairous hyper parameters and the independent variables of the model (5) Analyze the app’s screenshot images using deeplearning with tensorflow generating image tags along with their respective probabilities.