Ayswarya Sundararaman
Portfolio

Data professional specializing in Python, SQL, Tableau, data analysis, automation, machine learning, and data visualization.

@AyswaryaLinkedin

NETFLIX CASE STUDY

The Netflix case study focused on improving movie recommendations using machine learning algorithms like KNN, SVD, and XGBoost. Various models were tested, and RMSE was used as the evaluation metric. The case study aimed to enhance Netflix's recommendation accuracy by minimizing prediction errors, leading to significant improvements in prediction performance.

StackOverflow Tag Predictor

This project focuses on predicting tags for StackOverflow questions using machine learning models such as SGD Classifier, Logistic Regression, and MLkNN, leveraging TF-IDF and CountVectorizer for feature extraction and multilabel classification .

SGD Classifier Logistic Regression MLkNN Data Preprocessing Exploratory Data Analysis TF-IDF Multilabel Classification Python Scikit-learn

Walmart Store Sales Forecast

This project forecasts Walmart store sales using Random Forest, XGBoost, FB Prophet, and SARIMA, combining time series analysis with machine learning to improve predictive accuracy for holiday periods and promotions, the final model ranked 20th on Kaggle.

Random Forest XGBoost FB Prophet SARIMA Feature Engineering Time Series Analysis Python Scikit-learn Pandas

Taxi Demand Prediction

This project predicts taxi demand in New York City by analyzing taxi trip records from 2015 and 2016. The project segments pickup locations and predicts demand in different regions, improving transportation efficiency and resource allocation.

K-Means MiniBatch K-Means Random Forest XGBoost Feature Engineering Time Series Analysis Python Dask Seaborn

FB Friend Recommendation

This project predicts missing links in a social network graph for Facebook friend recommendations. It applies link prediction algorithms on a directed graph of Facebook connections to recommend potential friendships​

Jaccard Similarity Cosine Similarity PageRank Katz Centrality SVD Features NetworkX Python Scikit-learn

Microsoft Malware Detection

This case study involves extracting image features from ASM files and byte-level n-gram features to classify malware binaries. Machine learning models like Random Forest and XGBoost were used to improve malware detection accuracy. Feature selection techniques and visualization methods were applied to optimize prediction performance in identifying malicious software​

Bigram Features Image Feature Extraction Random Forest XGBoost Multithreading Python Scikit-learn

Quora Question Pair Similarity

This project identifies duplicate questions on Quora by analyzing question pairs to predict whether they are similar. It applies NLP techniques and advanced feature engineering. Various machine learning models like XGBoost and Logistic Regression are used to classify the question pairs.

TF-IDF Fuzzy Matching Stopword Removal Levenshtein Distance XGBoost Logistic Regression Feature Engineering Python Scikit-learn