VANDANAPU SRIHARI

Aspiring Data Scientist with strong mathematics background having knowledge on Python, Machine learning, Deep learning, Natural Language Processing, Market Basket Analysis and as well as time series forecasting. A punctual and hard working student, very interested in learning new stuff and research further to unknown information. Looking forword to adopt and learn from training in the new environment, also meeting new friends and to work together with them as a team.

Machine Learning       Bankruptcy Prediction

• The objective of the analysis is to predict that the company goes bankruptcy or not considering the different features of the company
• Analyzed and understood each of technical details related to business case, done EDA and model using python. Explained the data insights, outcomes, and experimental findings with teams.
• Used different classification algorithms, finally we have achieved 100% test accuracy and 100% train accuracy in Random Forest classifier after tuning the hyper parameters and deployed the model using Stream lit.

NLP             Hotel Review Classification

• The objective of the analysis is to predict that the customer has been satisfied or not and the rating based on review. Model gives Positive if the expected rating is >=3 out of 5 or else Negative.
• The text data is preprocessed like removing punctuations, numbers, extra white spaces is done before performing EDA. Analyzed the data with several EDA methods and tf-idf vectorization of text is done to perform the model building
• Several classification techniques are used to classify the ratings and finalized logistic regression model i.e having train accuracy of 94.7% and test accuracy of 92.6% and deployed the model using Stream lit.

Deep Learning     Image Classification using CNN

• The objective of the study is to predict images of 10 different classes correctly
• The size of the image is 32 x 32 pixels and the samples present are 60,000. The dataset used is CIFAR-10
• Built a CNN model with multiple convolution layers and multiple max pooling layers
• Built a model to classify images with 88% accuracy by performing batch normalization after every convolution layer

NLP       Web Scrapping and Sentiment analysis

• The objective of the analysis is to gather information about a product from the amazon website and use the text data for the sentiment analysis
• Used to identify accurately the people opinion from a large number of unstructured review text and classify them into sentiment classes
• Data preprocessing is done and followed by EDA and a Word cloud is build to understand the most repeating words in the text