COMPARATIVE ANALYSIS OF MACHINE LEARNING MODELS

Abstract

Data in its raw form might not mean much but after processing the data and making it more uniform it might reveal a lot of information. By using different types of machine learning algorithms, we can draw a lot of insights. This practice is already being carried out on a very large scale in today’s world but as the field of Machine Learning and Artificial Intelligence has advanced a lot, we have so many different algorithms at our disposal but the problem is data can be of many different types and there is no one algorithm that fits the best in every case. Using a complex model might not be useful for a simple dataset or vice versa and this practice might cost a company a lot of time, money and even after that the results might not be the best. Our goal is to depict this and identify which type of Algorithm gives the highest accuracy for which type of Dataset and identify the key factors that influence these algorithms, to demonstrate this we are using IRIS dataset and Wine quality dataset. Based on our research, we conclude that for simple and evenly distributed datasets such as Iris dataset, algorithms like KNN give the best results (95.5% accuracy). For non-uniform simple datasets such as Wine Quality dataset, algorithms like Decision Tree give 100% accuracy and KNN gives the lowest, 82.29%.