Table 4 Comparison between different knowledge extraction techniques. Stream processing applies different techniques to extract knowledge from arriving data. The techniques that best fit the envisioned mining scenario depends on the available data, on the training model, and on the expected goals of the processing
Technique | Data collection | Computing resource demand | Concept rift | Training | Supervised or unsupervised | Process | Typical applications | Related work |
|---|---|---|---|---|---|---|---|---|
Mining | Sampling and Summarizing | Usually low | No | No | Unsupervised | Preprocessing | Knowledge extraction from historical databases | |
Online learning | Feature extraction | Medium | Yes | Yes | Supervised and Unsupervised | Classification, regression, or aggregation | Knowledge extraction from online transactions | Andreoni Lopez et al. [10], Lobato et al. [79], Hultenn et al. [83], Lu et al. [85], Masuyama et al. [86], Polikar [87], Polikar et al. [88], Aaron et al. [89] |
Reinforcement learning | Model Free | Medium | Yes | Yes | Unsupervised | Reinforcement | Performance of controlling actions | Li et al. [90], Liu and Yoo [91], Tabrizi et al. [92], Chen and Qiu [93], Santos Filho et al. [94] |
Deep Learning | Implicit feature extraction | High | Yes | Yes | Supervised and Unsupervised | Classification, regression, or aggregation | Image processing | Deng and Yu [95], Kwon et al. [96], Deng [97], Kim et al. [98], Sze et al. [99], Wang et al. [101], Wang et al. [102], Turgut et al. [103], Wang et al. [104] |