In the last Post, the explanation about machine learning and what is descriptive analysis has been provided. In this post, I am going to provide some overview of the Predictive and Prescriptive analysis
Another analysis in machine learning in predictive analysis. Predictive analysis is about supervised learning. That means we want to learn from past data to predict the future trends. Predictive analysis forecast future based on some probability. There is no way to predict future with 100% probability. Predictive analysis mainly based on the probability of the statistic. We can classify predictive analysis into two main approaches:
In regression approaches, we predict a value. For instance, we may predict the sales for the next three months, or predict the number of website followers in the next three year.
While classification is about the predicting the group. Classification is about to predict the new item belong to which categories. For instance, in customer churn problem, we are going to predict a new custom belong to a group that may leave us or stays with us.
For both regression and classification, there are many algorithms that able to help us.
For regression algorithm, we have different algorithms such as Linear Regression, Logistic Regression, Polynomial Regression, Stepwise Regression, Ridge Regression, Lasso Regression, and ElasticNet Regression. Moreover, some other techniques such as Neural Network, decision tree and so forth.
For classification, there are many approaches such as SVM, Decision tree algorithms, KNN and so forth.Time series Forecasting is a subdomain of the predictive analysis and has some similarity with regression algorithms. Time series analysis forecast values based on time, time series data may have a trend or/and seasonality. Time series forecasts values based the past data trend, seasonality and existing residuals.
The prescriptive analysis is about the recommendation system. Recommendation system predicts the rating or preference users would give to an item. Recommendation system can be categorized into two main class:
· Content-Based System is the finding similarity between items. Content-based recommendation analyze item descriptions. And recommend an item that has more similarity to the user interest. For instance, we want to recommend new movies to a person, content-based filtering check the similarity of the new movies with other movies that the person already watched.
- Collaborative-filtering recommends items based on how similar users liked the items. Collaborative filtering recommends items based on how similar users liked the item