Abstract-Currently, massive data has been compiled and examined at explosive volumes. The vast volume of data stream structured by current technological sources in up-to-the-minute society has amplified at an incredible pace, stimulating processing capacity and data curation. The traditional eventdriven simulation models to investigate huge data can be no longer used but the paradigm to decisionmaking is now taken into all activities in the society, business applications, and scientific accounts. The overhead of data collection, redundant storage and processing improvement cost are on the rise for big data applications. In addition, there are technical aspects such as data inconsistency, redundancy, privacy, scalability, time-series and unusefulness [1]. Parallel processing is an essential method particularly for quantifying scale and time-series experiments. Simulation results for several datasets fit prediction results. Speedup and cost effective analysis will be considered as performance metrics as well.Keyword-Big Data Analytics, Massive Online Analysis (MOA), Parallel Processing, Performance Evaluation, Prediction I. INTRODUCTION The enormous volumes of data created by up-to-date high technologies in recent society introducing complications in processing cost and data classifications. On the other hand, analytical tools for cleansing and normalizing data streams are incapable to execute on real-time basis such explosive data size [5]. In addition, big data manipulation may exhibit volatile memory-paging, spatial locality and data flow control. On the other hand, the advent of parallel processing has disclosed the extraordinary processing power of PUs (processing units) to speed up data-explosive calculation much more as period elapses by. This research provides the evaluation of parallel processing for big data curation by using MOA (Massive Online Analysis) [6]. Another simulation tool, QuickSort, to evaluate genetic applications with parallelism has been proposed in [11]. Before executing experiments that implicate big data with parallelism, it is critical to split data sources into n subtasks. A previous research [5] exposes pros and cons of applying the graphic processors for computation. However, it is not resourceful unless the experiments are taking both partitioning and re-assembling time into consideration. Thus, in this research, analytical performance evaluation and prediction model are proposed by aiming the parallel processing architecture of MOA simulation as well as accounting both partitioning and re-assembling time. This paper is structured as follows. Firstly, section two outlines parallel big data description. Section three explains prediction method then section four describes simulation results and analysis. Lastly, section five draws conclusion and remarks the future work.II. OVERVIEW OF PARALLEL BIG DATA A deep learning is constructed up of machine learning's processors. A parallel processing [2] is made up for supporting several independent processing units accordingly. Clustering an...