An Ensemble-based Machine Learning Model for Accurate Predictions using Multiple Categorical Datasets

Abstract

In the consumer sector, electronic reviews are more common and comprehensive. Manufacturers, retailers, and customers are all aware of it. All require this knowledge to benefit from considering and guiding the massive and energetic data spaces that follow. Many social media outlets give polarized feedback. A fundamental problem with the internet's destructive content is that it makes it impossible for people to read important information. We'll look at all of the traditional machine learning approaches to catch the true sentiment. The accuracy using decision tree, naïve bayes and KNN applied on nursery dataset. All these three techniques achieved precisons of 72 to 90%. Decision tree performed well and accurate result as compared to knn and naïve bayes.The decision tree faced overfitting issues and KNN faced issues in deciding the value of K. On a large dataset, complexity increase when we apply a decision tree. Zero probability issues are a major challenge in naïve Bayes. To solve all those issues, an ensemble machine learning model (NKD) for accurate predictions is proposed to check the performance of the model. The proposed methodology applied on nursery dataset and IRIS dataset. The accuracy achieved using iris dataset is 100%.