Evolutionary algorithms have been successfully employed to solve classification problems. However, what I believe that hybridization of evolutionary approach have better precision rate as compared to their individual use. Moreover, other factors like nature of data, features selection, pre and post processing should also be considered
Most evolutionary algorithms are good so far for classification problem. In general, it depends on researcher view point and justification as well as the datasets that work on it
Most evolutionary algorithms can be used to solve the classification problem, which like use evolutionary algorithms to solve clustering problem. If the extended model be used, most classification accuracies are 100% on datasets. Let me know if you have interests.
As we know classification problems are of two types:
1. Supervised Learning (Classes are known)
2. Unsupervised Learning (Classes are unknown)
Now the question before us is:
1. How to use EAs in classification problem?
Now let's discuss the answer briefly
1. How to use EAs in classification problem?
A. Supervised Learning: In supervised learning EA is used either to tune some parameters of classification model (Support Vector Machine, Naive Bayes, K-Nearest Neighbour, etc.) or to select optimal number of features to achieve desire level of accuracy in less amount of time.
(i) The main purpose of tuning the classification model is to optimize the decision boundary for better learning and classification.
(ii) The major problem with the classification model is over fitting or curse of dimensionality. This problem arises due to the presence of redundant or repeatitive features. Therefore, EA helps in finding the optimal set of features which can improve the classification accuracy as well as significantly reduces the learning and classification time.
(iii) During experiment we also observed that these approaches can greatly impact the accuracy.
B. Unsuprevised Learning: Unsupervised learning is more of clustering approach in which the data are clustered on the
basis of similarity. Kmeans, Fuzzy C-means etc. are the most common and popular clustering approaches. Therefore, here
EA can be used to cluster the data according to the user defined objective.
Latest research published in PLOS and Nature on the same problem
Evolutionary algorithms have been successfully employed to solve classification problems. However, what I believe that hybridization of evolutionary approach have better precision rate as compared to their individual use. Moreover, other factors like nature of data, features selection, pre and post processing should also be considered