Hand-On Workshop on Time Series Modelling

On a vibrant August 8, 9, and 10, Christ College Pune became the hub of intellectual curiosity as it hosted a workshop on time series modeling, led by the esteemed data scientist Dr. Bindu Krishnan. The event drew a diverse audience, including students, faculty members, and industry professionals, all keen to delve into the intricacies of time series data and its wide-ranging applications.

Dr. Bindu Krishnan, a Doctorate in Statistics, has over two decades of experience in academics and research. Currently, she serves as a Senior Statistician in the Data & AI division at IBM Research and Development Software Labs, Kochi, Kerala. Her impressive career also includes her role as Professor and Head of the Department of Data Science, Computer Science & IT at Jain University, Kochi, and various other academic positions across prestigious institutions.

The workshop commenced with Dr. Krishnan providing an engaging introduction to time series modeling—a statistical technique crucial for analyzing data points collected or recorded at successive time intervals. She emphasized the significance of understanding time series data, which is pivotal for predicting future trends, understanding past behaviors, and making informed decisions across multiple domains, including finance, environmental science, and public health.

 

Dr. Krishnan introduced the fundamental components of time series data: trend, seasonality, and noise. She explained how the trend represents the long-term direction in the data, while seasonality captures regular, repeating patterns often influenced by factors such as seasons or holidays. Noise, she noted, refers to random variations that do not follow a discernible pattern. Through interactive examples and real-world datasets, participants learned to decompose time series data into these components, a process essential for effective analysis.

One of the key sections of the workshop focused on exploratory data analysis (EDA), an initial step in time series modeling. Dr. Krishnan demonstrated various techniques for visualizing and understanding time series data, such as line plots, histograms, and autocorrelation plots. She underscored the importance of these visual tools in identifying underlying patterns and anomalies before applying more advanced modeling techniques.

Participants were also introduced to powerful tools and libraries used in R, including functions for data cleaning, preprocessing, and basic statistical analysis. Dr. Krishnan guided them through the process of handling missing values and performing preliminary analyses, providing practical experience in preparing their data for modeling.

The core of the workshop delved into time series modeling techniques, with Dr. Krishnan covering several key models using R:

1. Autoregressive Integrated Moving Average (ARIMA): Dr. Krishnan explained how the ARIMA model combines autoregressive (AR) terms, differencing (I), and moving average (MA) terms to model time series data. She showed participants how to identify the appropriate order for these components using autocorrelation function (ACF) and partial autocorrelation function (PACF) plots.

2. Exponential Smoothing Methods: Dr. Krishnan also covered exponential smoothing methods, including Simple Exponential Smoothing (SES) and Holt-Winters Exponential Smoothing. These methods, known for their usefulness in short-term forecasts, are based on weighted averages of past observations.

Throughout the workshop, Dr. Krishnan stressed the importance of evaluating and validating time series models to ensure their accuracy and reliability. She introduced participants to various metrics for model evaluation, including Mean Absolute Error (MAE), Mean Squared Error (MSE), and Root Mean Squared Error (RMSE). These metrics, she explained, are vital for comparing different models and selecting the most suitable one for specific applications.

The workshop concluded with several hands-on exercises and case studies, allowing participants to apply the concepts and techniques they had learned. Using provided datasets and R, attendees built and evaluated different time series models, gaining valuable practical experience in the modeling process.

In summary, Dr. Bindu Krishnan’s workshop at Christ College Pune was an enlightening experience that provided participants with a deep understanding of time series modeling using R. The event not only broadened their knowledge but also equipped them with practical skills essential for analyzing and forecasting time series data across various fields.