Meena Muthusubramanian

In this project we explored job openings in the data science field that were listed on Glassdoor. We wanted to analyze data science job postings to see what kind of roles were most needed, where and in what industries these jobs were needed, what salary ranges were offered for different positions and what kinds of skills were companies looking for in candidates.

We started our research looking at different roles within the field of data science. Next, we looked at the 5 industries with the most data science roles in our data set and followed with an analysis of Glassdoor ratings and counts for companies with data science roles. We then looked at salary ranges for data science roles within different metro areas, for different positions and in the top 5 industries. Lastly, we peformed text analysis to see what kind of skills a candidate should be improving upon in looking for data science roles.

This application is developed for employees working in the Austin Animal Center. They can save animals' traits information like type, breed, color and date of birth. Also, the application allows users to save animals’ intake and outcome history. Users can not only insert new data, but also update and delete the data. Whenever users update the animal’s information, the application saves the new and old data together in the auditing table. The purpose of this function is that they can find old data easily when they update wrong data by mistake.

Our primary goal was to determine the attributes by exploring the characteristics of companies and to better understand the risks through analysing the incomplete payments. We figured out attributes which lead to won opportunities and also came up with major findings as listed :

• Current advertising efforts aren't maximizing the won oliortunites
• MBG should find ways to update the facebook marketing parameters to reflect certain attributes like industry and base wrt won and paid in full opportunities.

• To determine the top 100,000 customers that are likely to purchase tires in November of 2018.

We find that there are a few factors that are most important when predicting tire sales: days since last service and total sales on both services and tires. This is the notebook that contains our final model - a random forest model - along with all of our predicted IDs, and our process for obtaining them. Prior to creating this model, we also created a base model to get a baseline accuracy for comparison. We also needed to deal with the class imbalance and save this data for use in the future, found here. Finally, we tested out a few other models such as XGBoost and more model tuning to see if we could improve upon the accuracy obtained with the random forest model