Arpita Gupta

Developed risk profiling model for companies using unsupervised clustering techniques such as K prototypes, K means and hierarchical clustering. Applied web scraping using BeautifulSoup and urllib3 and performed filtering using a given set of keywords as well as cosine similarity. Implemented supervised sentiment analysis using Natural Language Processing and deep learning techniques such as CNN+ Word Embedding and LSTM achieving 90 per cent accuracy in the process. Devised text summarization model using KeyBERT and context-based word cloud using uniform distribution for scoring as well as frequency-based word cloud. Performed modelling for fraud detection using unsupervised techniques such as Isolation Forest, Local Outlier Factor and One Class SVM, improving the accuracy to 90 per cent in the process. Assisted in developing a supervised classification model for confidence index rating using Random Forest, which would speed up the time from three days to a day in accepting or rejecting names of company in a form.

Formulated dashboards using Excel and Power BI, and performance reports using SQL on a regular basis to identify performance and customer issues for a health insurance client bringing the repeat caller rate down to 5 per cent. Improved the efficiency of SQL queries for healthcare project by reducing the wait times by 66 per cent.