People sometimes ask me what are future trends in data mining. This was by the way the topic of an older post. I recently read a paper on this topic by Kriegel et al. (1). As clearly stated in the paper title – Future trends in data mining – this work points out future directions in data mining. After basic definitions about data mining and knowledge discovery, authors give examples of current and future issues in the field.
They mention several data mining challenges such as:
- Complex data types
- Incorporation of domain knowledge
- Distributed data mining
- Temporal aspects hidden in data
- Link between data preprocessing and data mining
- Automated data preprocessing
- Data integration
- Automated tuning
Two points I find promising are data integration and automated tuning. In the first, the aim is to avoid discovering patterns that are not caused by the phenomena we want to study. In the second, the goal is to reduce the time needed to find the best method and good hyper-parameters that give the best results. One example is the solution proposed by KXEN, where the data mining process is automated.
(1) Future trends in data mining, Kriegel et al., Journal of Data Mining and Knowledge Discovery, 15, pp. 87-97, 2007.