Fig. 2From: Design and analysis of a general vector space model for data classification in Internet of ThingsMacro precision for different combinations of methods [41]. The best dimensions for the different feature selection methods were different (including IG, MI, ECE, CDF). The figure shows that feature dimensions had a significant effect on the classification accuracy. With an increase in the feature dimensions, the macro precision of the classification algorithm also increased, but after reaching a certain threshold, the macro precision decreased instead. The best dimension for the IG method and the MI method was 450, that for ECE was 400, and that for CDF was 350Back to article page