Data mining is the process of extracting and discovering structure in huge datasets. It employs methods of machine learning, statistics and database systems to achieve the same. It is the science that turns raw data into useful information. Useful data is extracted employing intelligent methods from a huge dataset. It transforms the data into a comprehensible form for future use.  Knowledge discovery in databases or KKD is the analysis step of data mining. Database and data management, data pre-processing, model and interference considerations, interestingness metrics, complexity analysis, post-processing of the structures, visualizations and online updation are other aspects besides raw analysis. 

Data mining is the process of inferring the data from raw information, it is derived from huge amounts of data and not the extraction of the data itself. It is applied to any form of huge amounts of data and information processing like collection, extraction, warehousing, analysis and statistics. It sees the application of computer decision support systems, artificial intelligence like machine learning and business intelligence. 

The data mining process is actually a semi-automated or automatic analysis of huge amounts of data to derive previously unknown, interesting structures like groups of data records or cluster analysis, unusual records like anomaly detection and dependencies that are association rule mining and sequential pattern mining. This employs database techniques like spatial indices. These structures can then be visualized as a summary of input data. It can be further used in analysis, for example, in predictive analysis. The data mining will identify many groups in the raw data. This is used to gain precise results by the decision support system. Data collection, data preparation, result interpretation and reporting does not belong to the data mining process. They fall under the overall KDD process as additional processes. 

Data mining techniques use machine learning and statistics to reveal hidden patterns in large data, while data analysis is used to test models like analysis of the effectiveness of the data despite its volume. Data dredging, data fishing and data snooping makes use of data mining techniques to sample parts of a large volume of dataset that may be very small for realistic statistical inferences to validate any patterns discovered. All these techniques can be used to create new hypotheses to test large volumes of datasets. 

Information discovery in KDD process involves:

• Pre-processing. Targets must be made before using data mining algorithms. The targets set should be large so that it can contain the patterns while remaining aware enough to accomplish in stipulated time. This is because data mining can only uncover patterns that are already present in the dataset. Data warehouses or data mart are the sources for the data. Pre-processing is a very important step in data mining to analyze the multivariate datasets prior to data mining. The target set is then cleared. Data cleaning filters out the observations with noise and the component with missing data. 
• Data mining. Data mining comprises of 6 classes of tasks:

1. Anomaly detection. It is also called outlier and deviation detection. Unusual data records, data errors that need further studying are identified in anomaly detection. 
2. Association rule learning. It is also called dependency modeling. It usually searches for relationships between variables. Market basket analysis can be an example to this. A supermarket holds the information of its customers' purchase history. It analyses what items are usually bought together. This information is further used for marketing purposes. 
3. Clustering. It identifies the groups and structures in the data that are more or less similar without using known structures in the data. 
4. Classification. It generalizes the known structures to apply them on the other data. For example, the e-mail application attempts to segregate the mails as legitimate and spam and place them in separate folders. 
5. Regression. It finds the function to model the data with least error possible to estimate the relationship between the data or the dataset. 
6. Summarization. It gives a better compact representation of the dataset that includes visualization and report generation. 

• Result validation. It is possible to misuse data mining techniques unintentionally, but can still produce significant results. This cannot predict future behavior and cannot be applied on other samples of data. It is of very little use. This is the result of investing too many hypotheses and poor statistical hypothesis testing. This is called overfitting in machine learning terms. All the patterns that are found by data mining algorithms need not be necessarily valid. 

Test sets of data are used in evaluation for untrained data mining algorithms. Known patterns are applied to the test set. The resulting output is then compared to the expected output. If the known patterns do not match with the expected results, it must be re-evaluated. Also is it necessary to change pre-processing and data mining algorithms or to alter a few values. The final process has to be interpreted and turned into useful information.

Industry works with six phases of data mining:

• Business understanding
• Data understanding
• Data preparation
• Modeling 
• Evaluation
• Deployment

These processes are simplified into three simple processes: Pre-processes, Data mining and Result validation.

The free Data Mining course offered by Great Learning will help subscribers learn the subject for free online. The course will take you through various processes and phases involved in data mining. It takes you through various algorithms, selection, pre-processing, data-mining techniques and result validation processes in detail. You will also be taken through examples and applications of data mining. You can also learn free data mining tutorials in your free time. You will also earn a free certificate after the successful completion of the course. Happy learning!