Karan Nandankar

Research Intern

Tata Consultancy Services•  May 2019 - July 2019

• Developed a template for Time Series Forecasting and Anomaly detection in ordered and unordered series as main objective. • Reduced the computation time by 38%, by analysis of yearly &monthly behaviour of Global Superstore by initiating Spark RDD's • Examined the Stationarity by Dickey-Fuller Test and Rolling Statistics; stablised it by eliminating the Trend and Seasonality • Attained the RMSE of 153.2 by estimating the orders of ARIMA using the GridSearch and autocorrelation plots, ACF and PACF • Implemented Anomaly detection algorithms such as OneClassSVM, Isolation Forest and EllipticEnvelope for unordered series • Enhanced the F beta-score from 0.72 to 0.81 by performing LSTM based Enc-Dec for Anomaly detection in multisensor timeseries