Retrieval of Interactive requirements for Data Intensive Applications using Random Forest Classifier

Abstract

Classifying requirements in data-intensive systems based on their interactions can assist the requirements engineering process in becoming more systematic and transparent, resulting in higher requirement compliance and software project completion. However, understanding the requirements centred on interactions with the system is particularly tough due to the increased complexity of big data. In most cases, awareness of interaction-based requirements is critical in moving forward with prediction and decision-making. As a result, the classification of interactive requirements plays a critical role in removing the difficulties from unclear requirements. Various approaches to effective requirement classification are being devised. However, due to inadequate requirement management reflecting the fast-changing organizational change, classification accuracy does not achieve its maximum potential. The best approach for reducing misclassification rate and retrieving interactive requirements for data-intensive systems would be to use Word Embedding and a Fast Similarity Search (k-NN) retrieval mechanism, as none of the studies to date have emphasized it. It also assessed the impact by comparing the results to metrics derived from the Random Forest classifier's training on word count characteristics. The data set used to experiment with the classification, particularly for interaction-based needs, is unique to our work and has not been covered by any other studies to date. The researchers will benefit from this study as they will better understand the requirement classification process. With an F1 score of 0.91, precision of 0.89, and recall of 0.93, statistical analysis showed that Word Embedding followed by k-NN similarity search produced a relatively high classification result to differentiate interactive requirements for data-intensive systems.