vivekmehndiratta
Oct 9, 2019
Graduate / Description of methodologies and results while solving a problem - Texas Mccombs MSBA Essay [2]
Essay Question: Imagine that you are in a job interview. The interviewer asks you to describe a time that you have had to solve a problem, and what the results were. How would you answer? (If applicable, please describe the quantitative methods you employed.) (250 words maximum)
One of the most challenging problems I faced while working with one of the leading news clients was to increase their revenue by increasing traffic on their website. We decided to build a recommendation engine. While going through their web page, I realized an ample amount of space on the sides was left unused. I took this as an opportunity and filled it with news recommendations. The richness in recommendations will allow viewers to spend more time on the website which will drive more revenue.
I started by designing a content-based recommendation engine using topic modeling techniques. With approx. 120 articles published each day, keeping in mind the freshness of articles, I chose the last 10 days of articles for recommendations and the last 30 days of articles for training. I used information retrieval techniques and manipulated text data via NLP tools and trained the structured data using the NMF clustering which divided the bag of words into two matrices. The elbow method was used to determine the right number of cohorts. I used collaborative filtering to generate content-based recommendations. In the next step, hybrid-filtering was used to provide user-specific recommendations by involving the user history of articles.
After testing the model for 30 days, the results were extremely satisfying, an average viewer spent 25% more time and the customer retention rate was increased by 35%. The company could increase its revenue by 10%. We deployed this project for regional language news clients as well.
founding solution for a problem
Essay Question: Imagine that you are in a job interview. The interviewer asks you to describe a time that you have had to solve a problem, and what the results were. How would you answer? (If applicable, please describe the quantitative methods you employed.) (250 words maximum)
One of the most challenging problems I faced while working with one of the leading news clients was to increase their revenue by increasing traffic on their website. We decided to build a recommendation engine. While going through their web page, I realized an ample amount of space on the sides was left unused. I took this as an opportunity and filled it with news recommendations. The richness in recommendations will allow viewers to spend more time on the website which will drive more revenue.
I started by designing a content-based recommendation engine using topic modeling techniques. With approx. 120 articles published each day, keeping in mind the freshness of articles, I chose the last 10 days of articles for recommendations and the last 30 days of articles for training. I used information retrieval techniques and manipulated text data via NLP tools and trained the structured data using the NMF clustering which divided the bag of words into two matrices. The elbow method was used to determine the right number of cohorts. I used collaborative filtering to generate content-based recommendations. In the next step, hybrid-filtering was used to provide user-specific recommendations by involving the user history of articles.
After testing the model for 30 days, the results were extremely satisfying, an average viewer spent 25% more time and the customer retention rate was increased by 35%. The company could increase its revenue by 10%. We deployed this project for regional language news clients as well.
