Decision Tree-Based Data Mining and Rule Induction for Identifying High Quality Groundwater Zones to Water Supply Management: a Novel Hybrid Use of Data Mining and GIS

Abstract

Groundwater is an important source to supply drinking water demands in both arid and semiarid regions. Nevertheless, locating high quality drinking water is a major challenge in such areas. Against this background, this study proceeds to utilize and compare five decision treebased data mining algorithms including Ordinary Decision Tree (ODT), Random Forest (RF), Random Tree (RT), Chi-square Automatic Interaction Detector (CHAID), and Iterative Dichotomiser 3 (ID3) for rule induction in order to identify high quality groundwater zones for drinking purposes. The proposed methodology works by initially extracting key relevant variables affecting water quality (electrical conductivity, pH, hardness and chloride) out of a total of eight existing parameters, and using them as inputs for the rule induction process. The algorithms were evaluated with reference to both continuous and discrete datasets. The findings were speculative of the superiority, performance-wise, of rule induction using the continuous dataset as opposed to the discrete dataset. Based on validation results, in continuous dataset, RF and ODT showed higher and RT showed acceptable performance. The groundwater quality maps were generated by combining the effective parameters distribution maps using inducted rules from RF, ODT, and RT, in GIS environment. A quick glance at the generated maps reveals a drop in the quality of groundwater from south to north as well as from east to west in the study area. The RF showed the highest performance (accuracy of 97.10%) among its counterparts; and so the generated map based on rules inducted from RF is more reliable. The RF and ODT methods are more suitable in the case of continuous dataset and can be applied for rule induction to determine water quality with higher accuracy compared to other tested algorithms.

A Novel Approach to Harmonize Vulnerability Assessment in Carbonate and Detrital Aquifers at Basin Scale

“…Although many data mining techniques have been applied in groundwater vulnerability assessments [39,40,42], only a few studies have used decision trees in vulnerability studies [50,51]. They have been mainly used to assess other groundwater quality problems [65,67,68].…”

“…Decision trees can represent the relationship between variables and output class using specific rules following the pathway from the root node to the terminal node [63,65,66]. A priori, the number of terminal nodes on the tree can determine the number of rules but it may generate a large number of irrelevant pieces of information.…”

Classification and Prediction of Fecal Coliform in Stream Waters Using Decision Trees (DTs) for Upper Green River Watershed, Kentucky, USA

“…The arcsine transformation helps in making the land use distribution near-normal or Gaussian. This transformation gives the values in radians proportional to the angle subtended at the center on a pie chart of land uses [25]. The transformation is known to stabilize the variance and scales the proportional data [30][31][32][33].…”

“…Using the results, they were able to come up with the best ML model for sampling optimization. The other recent discussions and applications of Decision Trees in the context of water quality modeling in rivers include that of studies [19][20][21][22][23][24][25][26][27].…”

Comprehensive Review on Application of Machine Learning Algorithms for Water Quality Parameter Estimation Using Remote Sensing Data

“…Jeihouni et al used decision-tree-based data mining to identify high-quality groundwater zones for water supply management. (69) They used different DT methods such as ordinary decision tree (ODT), RF, random tree (RT), chi-square automatic interaction detector (CHAID), and iterative dichotomiser 3 (ID3) to extract key relevant variables affecting water quality (electrical conductivity, pH, hardness, and chloride) in a GIS platform. The RF showed the highest performance (accuracy of 97.10%) among the methods.…”