With the latest innovations, smart phones tend to be an essential part of everyday life used for managing several tasks. The major applications include IoT, Intelligent Vehicles, Cyber Physical Framework, Industry 4.0, Embedded Devices, etc. As the technology improves, multiple cores are integrated to convert the design to a Heterogeneous System-On-Chip (HSoC) design constrained for performance improvement. In real time applications, prediction of core for mapping workload and optimization of energy consumption in multi-core processors in HSoC remains a major challenge. The database contains characteristics of workload such as memory, branch data, instruction cycles, etc. The algorithms like Support Vector Machine (SVM), naive baize, random forest, KNN, Machine Algorithms, Deep Neural Networks (DNN), LSTM are used for prediction of best core for each workload during runtime. The Rasp-pi quad core processor is considered for simulation using python IDE with machine learning library, SPEC (CPU-2006), MiBench and IoMT Benchmarks. The metrics used for comparison are prediction accuracy, precision, energy consumption, throughput, selectivity, affectability etc. The accuracy in prediction was observed to be 98%, precision to be 96%, savings in execution energy up to 30%.