New Trends in Data Mining - Data Mining

Data mining is relatively young compared to database technology. It is still considered a niche and emerging market. One of the reasons for this designation is that most of the data mining packages are targeted toward statisticians or data miners. Application developers find it too challenging to master these technologies. More recently a number of data mining vendors, including Microsoft, realized this and initialized data mining APIs that are directed toward the developer communities. We believe that in the next few years there will be more and more developers who will be able to build mining models, and as a consequence a large number of applications will include data mining features.

We foresee the following trends in data mining area over the next few years:

Embedded data mining: More and more business applications will include data mining features, in particular the prediction feature, for added value. For example, the CRM application will allow users to forecast product sales. Online retailers will recommend products to customers for cross-selling purposes. This will be due largely to the fact that industrial data mining APIs, such as OLE DB for Data Mining, enable database and application developers to use data mining features and embed them in a line of business applications. Embedded data mining will increase the overall size of data mining market.

Data mining packages for vertical applications: Data mining is becoming popular as major database vendors add it to their database management system (DBMS) packages. Data mining can be applied to almost every sector. Today, major data mining markets exist in finance, insurance, and telecom. There is a growing need for specialized data mining techniques to solve business problems in many vertical sectors. For example, in the health care field, we need special data mining techniques to analyze DNA sequences. In network security applications, we need real time training algorithms to detect network intrusion. We need nontraditional data mining techniques to analyze the unstructured data in the World Wide Web. Text mining is yet another vertical sector to which we need to apply data mining. Traditional horizontal data mining packages are too general to solve these problems. We foresee there will be more new data mining packages specialized for these vertical sectors.

Products consolidation: Hundreds of software vendors are providing horizontal data mining packages. Many packages include only one or two algorithms. The data mining market is still very fragmented. Just as with other software sectors, consolidation is inevitable. Small vendors will find more competitive pressure in the horizontal market, especially when major database vendors add data mining features to DBMSs.

PMML: Although big vendors, such as Microsoft, Oracle, IBM, and SAS, are competing on various data mining APIs, they are all member of the same club: the DMG (Data Mining Group). They all support PMML as the model persistence format. PMML offers many advantages in terms of model exchange and model deployment. Because it is an XML document, it is also editable by advanced users. PMML will become more popular in the near future.

All rights reserved © 2020 Wisdom IT Services India Pvt. Ltd DMCA.com Protection Status

Data Mining Topics