Originally Posted On: https://exyte.com/blog/data-mining-how-different-types-of-modeling-work
Data mining is a system of analytical methods designed to deal with various organizational issue in order to discover patterns in large quantities of data.
Data mining is a complex system of analytical methods and techniques designed to deal with various organizational issues, create questions and rules in order to discover patterns in large quantities of data. This system can be roughly divided into descriptive, predictive, and prescriptive modeling.Descriptive modeling
Descriptive modeling includes data mining methods for grouping and categorizing data:
- Clustering to find groups and systems within the data – this is an easy and the most straightforward method;
- Outlier detection to identify irregular data or errors, which require further actions;
- Association rule learning used to establish connections between the elements of the data;
- A principal component analysis made to help establish if there are any correlations between the units of the data;
- Affinity grouping, which refers to forming groups of elements according to how similar they are.
This method is commonly applied when a company needs to summarize and describe various aspects of its operations.Predictive modeling
Predictive modeling includes data mining methods based on predicting how something is going to behave:
- A regression method estimates how different elements or clusters of information interact;
- Neural networks, computer software algorithms, which are able to identify patterns and predictions;
- Decision tree diagrams, which depict ‘branches’, each of them standing for different probabilities.
- Predictive analysis is utilized to gain a forecast of future events or fill up a gap in one’s knowledge about the upcoming events.
Prescriptive modeling includes data mining methods designed to determine the best course of action:
- Creating if/then rules designed to help you understand what can happen and what you should do in each possible scenario;
- Applying computational modeling to reveal optimal courses of action.
- This method is aimed at providing instructions regarding which actions should be taken to achieve the best results.