Synthetic Time Series Data Generation Using Time GAN with Synthetic and Real-Time Data Analysis

Abstract

Synthetic time series data generation is a wide area to research, and lot of attention has drawn recently. The assumption for generating multivariate time sequenced data is well proportioned and continuous without missing values. A productive time series data model should retain momentum, such that the new sequence maintains the actual relationship between different variations occurrence over time. Different existing approaches that put forward generative adversarial networks into the sequence system that prohibits care for temporary interactions that differ from sequential time data. Simultaneously, monitored sequence prediction models—allowing for fine control over network dynamics—are naturally determined. Time series data generation has problems like informative missing data values leading to untraceable challenge and long sequences with variable length. These problems are biggest challenges in making of a powerful generative algorithm. Herein, we are using an innovative structure to produce real-time sequence which associates the adaptability of different unattended prototype through controls provided during the supervised model training. Privacy data analysis safeguards data privacy and data sharing. Generation of accurate private synthetic data is a NP hard problem from any considered scenario. This paper discusses about privacy parameter and data analysis of real and synthetic data. The synthetic data is enclosed in the boundaries of real data and have similar behavior.