Data format figures-DATA MINING LEARNING MODELS AND ALGORITHMS ON A SCADA SYSTEM DATA REPOSITORY

The original data set included noisy, missing and inconsistent data. Data
preprocessing improved the quality of the data and facilitated e±cient data
mining tasks.
Before the experiment, we prepared data suitable to next operation as
following steps:
² Delete or replace missing values;
² Delete redundant properties (columns);
² Data Transformation;
² Data Discretization;
² Export data to a required .ar® or .csv format ¯le [11].
The original and modi¯ed formats of data set are shown in Figure 1 and
Figure 2.
Data visualization is also a very useful technique because it helps to deter-
mine the di±culty of the learning problem. We visualized with Weka single
attributes (1-d) and pairs of attributes (2-d). The ¯gure 3 shows the variation
of the temperature in time.