LATEST PROJECTS
Project | 01
Text Analysis | Python
Work in Progress: creating a text classifier model using deep learning techniques (including transfer learning and RNN) for predicting sentence type (question/statement/apology). Developed a sentence scorer model to score syntactic and semantic conversational compliance for Chabot using the Spacy library. Created a flask app for deploying the text classifier model as api on server. Build a UI screen using HTML for taking text input from the end-user and displaying its target class.
Project | 02
Clustering Analysis | R
Work Done:: The dataset had 323,849 records and 633 variables. This was students' data and clustering them will help the advertisement team create ads based on each cluster's persona.
Data Reduction: Reduce data missing with 10% values. Use business knowledge to keep important variables. Replace numerical data with mean. Fix categorical data with correct factors and impute the missing data. Apply the appropriate algorithm (cLUSTMIX in R) for doing clustering of data of mixed type.
Project | 03
Marketing Mix Model | R
Work Done: Marketing spends data from each media - TV, Radio, Billboard, and Digital for a period of time provided to us in reports format. Using those reports we extracted spend for each week from June 2017 to June 2018. Data of student inquiries for each of the weeks during the same period also provided to us. Our goal is to determine which marketing media is most effective for bringing inquiries.
Process: I used regression modeling techniques to determine the number of inquiries that came in for each week based on spending by each media. All the models concluded least p values for digital media and high values for other media. Thus concluding traditional media has a less direct impact in bringing inquiries compared to digital media.
Project | 04
Data Warehouse | Google Cloud
Work Done: Using Big query to load all of customer's CRM data, Google analytics data and ad spend data (from 3rd party API like supermetrics). Build comprehensive reports in google data studio to determine which advertisement brought which CRM lead.