A Novel Classification-Based Approach for Quicker Prediction of Redshift Using Apache Spark

Over the years observational astronomy is generating a huge amount of data due to the usage of high-resolution telescopic cameras. Extracting useful information from this huge data is an essential task in present days. Using this information, it will be easy to understand the universe better. In this paper, we have implemented four classification-based models such as Naive Bayes, Support Vector Machine, Decision Trees, and Random Forests using Apache Spark and done a detailed comparative analysis of those models in the redshift prediction task. Redshift is a metric used widely for distance measurement. Pyspark has been used for this task to process big data easily and quickly. Although redshift estimation is a regression task, we have proposed a new classification-based approach where redshift values have been divided into multiple bins and each bin is treated as a class. This approach helped us to achieve reasonably good accuracy while predicting redshift in the pyspark environment. Out of four models, random forest outperforms others in predicting redshift with the best accuracy of 95.05%.