Modern-day time series classification (TSC) in networks has emerged as famous and is presently an energetic research location. However, traditional methods in TSC in networks, including okay-nearest neighbors and help vector machines, require large training units and an extensive computational attempt as they need more capacity to extract complicated capabilities. This paper explores Multi-Layer Perceptrons (MLPs) as an alternative records-pushed approach to tackle the TSC task. We first design experiments to benchmark the performance of trendy MLPs educated on representative time series datasets from diverse domain names. We then look at the different architectures of state-of-the-art MLPs for the TSC hassle by exploring contemporary activation functions and structure variations. In addition, we speak about how distinct initialization techniques and hyperparameters may impact the schooling state-of-the-art MLP version. The outcomes reveal the higher accuracy of the MLPs in TSC networks. Eventually, we gift viable traces of the latest similar investigation for the proposed method. The technical summary explores using multi-layer perceptrons (MLP) for network time collection classification. MLP is a supervised deep mastering class set of rules used to expect time series lessons from entered data.