Towards Synthetic Data Generation Using Predictive Models

Abstract

Missing data are a common issue in real-world datasets and can significantly affect the performance of data analysis and machine learning tasks. Beyond their traditional role in estimating missing val ues, data imputation techniques can also be exploited for synthetic data generation. By learning the underlying relationships and statistical patterns within a dataset, predictive models such as Random Forest (RF) and Long Short-Term Memory (LSTM) networks can be adapted to generate realistic synthetic observations while preserving the characteristics of the original data. This Final Studies Project proposes an approach based on predictive modeling and artificial intel ligence techniques for both missing data imputation and synthetic data generation. The developed framework aims to estimate missing values, learn data distributions, and generate new data instances that maintain the statistical properties and variability of the original dataset. Such an approach can support data augmentation and synthetic data creation in domains where real data are scarce, sensitive, or expensive to collect.