Shashank Goswami

• Recommended EV charging station network distribution heat map across 1800 zip codes of California, performing Simpy simulation, yielding wait time with greatest ROI as an evaluation parameter.
• Segmented marketing targets by consumer profile using K-mean, PCA that can be used in marketing strategy.

• Collated data by scraping 4000 customer reviews from TripAdvisor using Selenium web driver in python, enabling data preparation.
• Cleansed the data by Stopwords removal, Stemming and Lemmatizing reviews using NLTK, shrinking text corpus by 30%.
• Programmed Sentiment analysis and Topic Modelling on customer reviews of 3-star hotels using vaderSentiment composite score, detecting their popular features.
• Mapped customer needs with hotel features and scored hotels according to… Show more

• Collated data by scraping 4000 customer reviews from TripAdvisor using Selenium web driver in python, enabling data preparation.
• Cleansed the data by Stopwords removal, Stemming and Lemmatizing reviews using NLTK, shrinking text corpus by 30%.
• Programmed Sentiment analysis and Topic Modelling on customer reviews of 3-star hotels using vaderSentiment composite score, detecting their popular features.
• Mapped customer needs with hotel features and scored hotels according to customer preference, thus recommending the hotel with the highest score. Show less

• Condensed JSON formatted multilabel text fields constituting 80% of the data utilizing python library “Abstract Syntax Trees”, optimizing data preprocessing.
• Transformed movie summary text using TFIDF into quantitative values, measuring the uniqueness of movie content.
• Revamped multilabel and multiclass fields by generating dummy variables deploying MultilabelBinarizer and OneHotEncoding, creating model ready inputs.
• Assessed impact of a Director, Actor, Budget, etc. on the… Show more

• Condensed JSON formatted multilabel text fields constituting 80% of the data utilizing python library “Abstract Syntax Trees”, optimizing data preprocessing.
• Transformed movie summary text using TFIDF into quantitative values, measuring the uniqueness of movie content.
• Revamped multilabel and multiclass fields by generating dummy variables deploying MultilabelBinarizer and OneHotEncoding, creating model ready inputs.
• Assessed impact of a Director, Actor, Budget, etc. on the movie revenue, employing sci-kit learn ExtraTreesClassifier, yielding 5 prime features contributing 60% to Randomforest prediction with evaluation parameter RMSE equals 1.58. Show less