| Title |
Prediction and Performance Evaluation of Coagulant in Water Purification Plant Based on Interval-Based Incremental Granular Model |
| Authors |
염찬욱(Chan-Uk Yeom) ; 곽근창(Keun-Chang Kwak) |
| DOI |
https://doi.org/10.5370/KIEEP.2021.70.2.102 |
| Keywords |
Incremental granular model; Interval-based fuzzy C-means clustering; Coverage; Specificity; Performance evaluation; Coagulant injection |
| Abstract |
In this paper, we design an interval-based incremental granular model(IGM) and compare and analyze the prediction performance according to the method of generating information granule in the output space. It is a structure that combines an incremental granular model with a linear regression(LR) model and a granular model(GM). The existing granular model uses the context-based fuzzy C-means(CFCM) clustering method, but in this paper, an interval-based granular model using the interval-based fuzzy C-means(IFCM) clustering method is used. Granular models are designed by information granules generated in the input and output spaces. The information granules generated in the input space are directly induced by the interval generated in the output space. In general, when creating an interval in the output space, a method of dividing the interval evenly and not overlapping is used. In this paper, in addition to the general segmentation method, we compare and analyze prediction performance by adding a method of flexibly and non-overlapping segments based on a probability distribution, and a method of equally dividing segments at regular intervals. Fuzzy-related prediction models are generally evaluated by RMSE and MAPE. However, in the case of an interval-based particle model, the final output is not a numerical value, but a fuzzy number value in the form of an interval, which is not suitable for evaluating the granular model. In order to evaluate the prediction performance of the interval-based incremental granular model, the performance index method using the coverage and specificity of the interval, which is the result of the granular model, is used.
In order to confirm the validity of the granular model, an experiment was conducted using predicted data for the injection of a coagulant at a water purification plant |