The organization uses data mining to improve decision making by discovering effective patterns and insights from the data.
To effectively mine data, huge amounts of data has to be analyzed and summarized so that it can be used to make logical decisions.
The results of data mining are usually data models which explain the underlying patterns and relationships. These models can be used for decision making which can be viewed with dashboards and reports.
Data mining can be used in both a supervised or an unsupervised analysis capacity. In the supervised analysis capacity, someone has to respond to a request while in unsupervised analysis, a pattern is detected and used in making automatic decisions.
Data can be mined in various ways such as describe descriptive, diagnostic, and predictive ways :
• Descriptive: this method identifies patterns in the data, for example. grouping customers based on the services which they use.
• Diagnostic: this is when techniques such as decision trees are used to identify why a pattern exists, for example the attributes of an organization’s most profitable customers.
• Predictive: this is when the mined data is used to predict future patterns for example the probability of a customer defaulting on a loan.
In order to make effective decisions when the the data mining work is being performed, the stakeholder needs have to be properly identified and communicated.
There are some components of data mining which are:
1. Requirements elicitation: The objective of data mining is to set up the decision requirements for a recognized business need. Data mining is especially useful in the agile environment because they could help with quick deployment and rework.
2. Data preparation: Data records are formed by merging data for numerous sources into a single data set. The data mining tools then work on these dataset.
3. Data Analysis: Once the data set is available, statistical measures are used to analyse the data and visual tools are used to view how these data values are dispersed, what data is missing, and the unique data characteristics.
for example: a characteristic could be the number of times a vendor delivered goods on time in the last 70 days. Determining that the count over the last 90 days is more useful than the count over the last 60 or 90 is important.
4. Modelling Techniques: There are numerous data mining techniques.
Which include:
• classification and regression trees (CART).
• linear and logistic regression.
• neural networks.
• support sector machines.
• predictive (additive) scorecards.
The dataset and its characteristics are inputted into these algorithms either automatically or manually.
5. Deployment: After the data model is built , it has to be deployed so that it can be used by the end user. The deployment can be automated or manual. Some data mining techniques such as the predictive analytic technique requires the use of business rules to ensure that they fulfil the requirements.
Data mining techniques has both its strengths and limitations which include:
Strengths
- It could identify hidden patterns and create useful analytical insights.
- It could identify useful data and predict future trends.
- It could be merged with the system design to improve data accuracy.
- It could remove human prejudice in decision making.
Limitations
- The data miner has to be experienced to accurately apply the techniques..
- Some of the techniques require the use of advanced math skills.
- Some stakeholders might be suspicious of the data mining techniques and its results due to its complexity.
- The stakeholders might then chose not deploy the data mining tool due to a lack of understanding.