Most of the study have used confusion metrics and performance measures like accuracy, recall, precision and f-score for presenting the results and comparing the machine learning methods. What is the significance of each of the methods and is their any alternate measures available for presenting the results and for concluding the better classifier?
As far as I know, accuracy is one of the basic metrics to measure classification success. Recall, precision and f-scores (and rand index) are more commonly used for clustering. In the classification, a specific sample should be clearly included in a class according to the attributes of its. So, accuracy controls if each sample in true class or not, and finds rate of rightness or error. However, in clustering, it deals with how samples in same cluster are actually similar to each other. Thus, similarity metrics are used to measure clustering success by using similarities or dissimilarities of samples.
The performance of the Machine Learning algorithms are evaluated based on the type of metrics used. The weblink gives u enough information about different performance metrics:
https://medium.com/greyatom/performance-metrics-for-classification-problems-in-machine-learning-part-i-b085d432082b
I think your question is broad as solving a problem in AI using ML depends on several aspects such as data, algorithms, programming proficiency and how do you want to present your results and which findings you want to be highlighted . In my humble opinion you need to read in depth in analytical methods. However learning about the tools that able to extract insight/value from data is important if you want to put a fingerprint in your specific research area.
- Badges
- Science topic