Computer Science and Engg

Project Title: Event Modelling and Predictive Analysis

Funding Agency: IOE FRP, Delhi University, Delhi
Amount: ₹3.5 L                                                                                               Year: 2020-21
Sanction Letter No and Date: lOE/FRP/PCMS/2020/27 dated 01.01.2021

Objectives of the Proposal:

1. To develop dynamic event modeling frameworks capable of capturing temporal dependencies and non-linear patterns in complex datasets.
2. To apply advanced statistical techniques (Bayesian inference, MCMC) and machine/deep learning models for predictive analysis across physiological, psychological, and social domains.
3. To design and implement scalable time-series models for early detection and prediction of critical events, such as emotional states, system reliability failures, and behavioral patterns.
4. To integrate feature optimization methods to enhance model precision and interpretability for diverse event-driven data structures, including EEG signals, lifetime distributions, and social media text.
5. To deliver real-time, high-accuracy predictive systems that support proactive decision-making in healthcare diagnostics, psychological profiling, and reliability engineering.

Innovations / Outcomes:

The project introduced multiple innovations across affective computing, reliability analysis, and personality analytics. A novel 3D emotion recognition system was developed for EEG signals using overlapping sliding window techniques, enhancing the classification performance on the AMIGOS dataset with 2D CNN models, achieving over 96% accuracy. The methodology addressed challenges of imbalanced and variable-length EEG data, presenting a first-of-its-kind robust framework.

In reliability analysis, Bayesian estimation and MCMC methods were applied to the Nakagami distribution under generalized Type-II progressive hybrid censoring, yielding highly precise parameter estimates validated through both simulated and real datasets. For social media analytics, personality prediction models were designed using machine learning algorithms integrated with advanced feature extraction techniques like TF-IDF and GloVe, significantly improving classification accuracy.

These outcomes contribute scalable, interpretable tools for emotion recognition, lifetime data modeling, and behavioral analytics, offering practical applications in healthcare diagnostics, reliability engineering, and psychological profiling.

Publications:

1. Pandey R, Ali D. Bayesian Analysis of Nakagami Distribution under Generalized Type-II Progressive Hybrid Censoring Scheme. Journal of Scientific Research. 2021;65(5).
2. Garg S, Garg A. Comparison of machine learning algorithms for content based personality resolution of tweets. Social Sciences & Humanities Open. 2021 Jan 1;4(1):100178.
3. Garg S, Patro RK, Behera S, Tigga NP, Pandey R. An overlapping sliding window and combined features based emotion recognition system for EEG signals. Applied Computing and Informatics. 2025 Jan 30;21(1/2):114-30.