Organizations today are gathering ever-growing volumes of information from all kinds of sources, including websites, enterprise applications, social media, mobile devices, and increasingly the internet of things (IoT).
The big question is: How can you derive real business value from this information? That’s where data mining can contribute in a big way. Data mining is the automated process of sorting through huge data sets to identify trends and patterns and establish relationships, to solve business problems or generate new opportunities through the analysis of the data.
It’s not just a matter of looking at data to see what has happened in the past to be able to act intelligently in the present. Data mining tools and techniques let you predict what’s going to happen in the future and act accordingly to take advantage of coming trends.
The term “data mining” is used quite broadly in the IT industry. It often applied to a variety of large-scale data-processing activities such as collecting, extracting, warehousing, and analyzing data. It can also encompass decision-support applications and technologies such as , , and business intelligence.
to help increase the loyalty of existing customers and attract new ones.
The key components of data mining
The process of data mining includes several distinct components that address different needs:
- Preprocessing. Before you can apply data mining algorithms, you need to build a target data set. One common source for data is a data mart or warehouse. You need to perform preprocessing to be able to analyze the data sets.
- Data cleansing and preparation. The target data set must be cleaned and otherwise prepared, to remove “noise,” address missing values, filter outlying data points (for anomaly detection) to remove errors or do further exploration, create segmentation rules, and perform other functions related to data preparation.
- Association rule learning (also known as market basket analysis). These tools search for relationships among variables in a data set, such as determining which products in a store are often purchased together.
- Clustering. This feature of data mining is used to discover groups and structures in data sets that are in some way similar to each other, without using known structures in the data.
- Classification. Tools that perform classification generalize known structures to apply to new data points, such as when an email application tries to classify a message as legitimate mail or spam.
- Regression. This data mining technique tis used to predict a range of numeric values, such as sales, housing values, temperatures, or prices when given a particular data set.
- Summarization. This technique provides a compact representation of a data set, including visualization and report generation.
Dozens of vendors provide data mining software tools, some offering proprietary software and others delivering products via open source efforts.
Among the key vendors that offer proprietary data-mining software applications are Angoss, Clarabridge, IBM, Microsoft, Open Text, Oracle, RapidMiner, SAS Institute, and SAP.
Organizations that provide open source data mining software and applications include Carrot2, Knime, Massive Online Analysis, ML-Flex, Orange, UIMA, and Weka.
The risks and challenges of data mining
Data mining comes with its share of risks and challenges. As with any technology that involves the use of potentially sensitive or personally identifiable information, security and privacy are among the biggest concerns.
At a fundamental level, the data being mined needs to be complete, accurate, and reliable; after all, you’re using it to make significant business decisions and often to interact with the public, regulators, investors, and business partners. Modern forms of data also require new kinds of technologies, such as for bringing together data sets from a variety of distributed computing environments (aka ) and for more complex data, such as images and video, temporal data, and spatial data.
Getting the right data and then pulling it together so it can be mined isn’t the end of the challenge for IT. The cloud, storage, and network systems need to enable high performance of the data mining tools. And the resulting information from the data mining needs to be presented clearly to the wide range of users expected to act on and interpret it. You’ll need people with and related areas.
, the idea of mining information that relates to how people behave, what they buy, what websites they visit, and so on can set off concerns about companies gathering too much information. That affects not just your technological implementation but your business strategy and risk profile.
Beyond the ethics of tracking individuals so thoroughly, there are also legal requirements about how data can be gathered, identified to a person, and shared. The United States’ and the European Union’s are among the best known.
In data mining, the initial act of preparation itself, such as aggregating and then rationalizing data, can disclose information or patterns the might compromise the confidentiality of the data. Thus, it’s possible to inadvertently run afoul of ethical concerns or legal requirements.
Data mining also requires data protection every step of the way, to make sure data is not stolen, altered, or accessed secretly. Security tools include encryption, access controls and network security mechanisms.
Data mining is a key differentiator
Despite these challenges, data mining has become a vital component of the IT strategies at many organizations that seek to gain value from all the information they’re gathering or can access. This drive will no doubt accelerate with ongoing advancements in predictive analytics, artificial intelligence, machine learning, and other related technologies.