• Vasavi Ravuri

Early event prediction in a longitudinal study is a significant and difficult problem with considerable practical significance in many real-world applications. In contrast to typical classification and regression problems where a domain expert may provide labels for the data quickly, training data in these longitudinal studies must only be acquired by waiting for the occurrence of a significant number of events. Utilizing data gathered in the past over a predetermined period of time, survival analysis seeks to directly anticipate the time to an event of interest. It cannot, however, provide a response to the unanswered query of "how to forecast whether a subject will experience an event by end of a longitudinal study using event occurrence information of other subjects at the early stage of the study?" The goal of this study is to predict an event's recurrence at a future time point using only information from a limited sample of events that occurred at the beginning of a longitudinal study. Due to the censoring of data on event occurrence and the availability of just a small number of data on events that occurred during the initial phase of the inquiry, this issue presents two important challenges. In order to create event prediction models that are trained early on in longitudinal research, we offer a novel Early-Stage Prediction (ESP) framework. First, using the Kaplan-Meier estimator, we create a new technique for dealing with censored data in order to address the first obstacle. We next build “three algorithms, namely, ESP-NB, ESP-TAN, and ESPBN, to efficiently forecast event occurrence utilizing training data obtained at an early stage of the investigation. These algorithms enhance the Naive Bayes, Tree-Augmented Naive Bayes (TAN), and Bayesian Network approaches based on the proposed framework”. More particularly, by modifying the prior probability of the event's occurrence for future time points, our method successfully combines Bayesian methodologies with an Accelerated Failure Time (AFT) model. With the aid of numerous synthetic and actual benchmark datasets, the suggested framework is assessed. Our extensive collection of trials shows that the suggested ESP framework is, on average, 20% more accurate than existing systems even using only a little quantity of event information in the training data.

"Bayesian approach for Predicting in longitudinal data at prima stage", International Journal of Emerging Technologies and Innovative Research (www.jetir.org), ISSN:2349-5162, Vol.10, Issue 8, page no.g111-g121, August-2023, Available :http://www.jetir.org/papers/JETIR2308612.pdf

"Bayesian approach for Predicting in longitudinal data at prima stage", International Journal of Emerging Technologies and Innovative Research (www.jetir.org | UGC and issn Approved), ISSN:2349-5162, Vol.10, Issue 8, page no. ppg111-g121, August-2023, Available at : http://www.jetir.org/papers/JETIR2308612.pdf