Internet of Things (IoT), Big Data (BD), Artificial Intelligence (AI) and Machine Learning (ML) are the novel approaches where communication happens between man-made machines. Machines interact and acquire knowledge by implementing learning algorithms. Data analytics, prediction and classification methods are machine learning approaches applied on Big data for processing various unstructured data patterns. MapReduce is a widely used programming framework to parallelize these machine learning algorithms. To accomplish best outcomes, the algorithms are fine tuned using parallel practice. This technique uses MapReduce model for processing datasets multiple times by tuning the parameters as per the requirement. But this existing MapReduce model endures with high disk rates resulting in low throughput and inefficient time complexity. To achieve the minimal time consumption for tuning the jobs, Apache Spark framework replaces the MapReduce model. This is examined in this paper by evaluating the prediction on "Demand and Supply of India" dataset. A comparative analytical study is proposed in this paper to predict the demand for forecasting by training the existing data using tree based machine learning techniques. The prediction outcomes computed are compared on tree structured ML methods with respect to time and space utilization.