KR20120136565A

KR20120136565A - Red tide blooms prediction method using fuzzy reasoning

Info

Publication number: KR20120136565A
Application number: KR1020110055570A
Authority: KR
Inventors: 박선; 임정수; 이성호; 김영주; 이성로; 양후열
Original assignee: 목포대학교산학협력단
Priority date: 2011-06-09
Filing date: 2011-06-09
Publication date: 2012-12-20

Abstract

Disclosed is a red tide occurrence prediction method using fuzzy inference. The red tide occurrence prediction method includes a clustering step of estimating and clustering a group having similar characteristics from a data set, calculating a fuzzy membership using a fuzzy membership function for the estimated cluster, and the calculated fuzzy membership. Generating a fuzzy rule for the input data, wherein the clustering step is characterized by using Chiu's method for estimating the cluster and using the cluster to estimate the potential center of the cluster. This increases the red tide occurrence prediction accuracy.

Description

Red tide blooms prediction method using fuzzy reasoning}

The present invention relates to a red tide related technology, and more particularly to a red tide occurrence prediction technology.

Fdez-rvierola et al. Proposed a prediction system for red tide. Their method uses a case-based method based on a hybrid of neural networks and fuzzy. A self-organized feature map (SOFM) neural network was used in the case retrieval phase, and a radial bassis function (RBF) neural network was used in the re-use phase. Fuzzy was used in the case correction stage, and fuzzy and RBF and self-organizing feature map neural networks were used in the case maintenance stage.

The red tide prediction monitoring system using this case-based reasoning creates a case base for inference and uses the KNN algorithm to search for the most similar cases. However, this method merely classifies whether the input data is red tide based on existing red tide occurrence cases.

An object of the present invention is to provide a red tide occurrence prediction method using fuzzy inference that can increase the red tide occurrence prediction accuracy.

The red tide occurrence prediction process according to the present invention includes a preprocessing and a fuzzy inference step. In the pretreatment stage, marine environment data of past red tide occurrences are processed into learning data suitable for fuzzy inference. In the fuzzy inference step, fuzzy inference rules are generated using learning data, and red tide occurrence is predicted using the generated rules.

Red tide occurrence prediction method using a fuzzy inference according to an aspect of the present invention for achieving the above technical problem is a clustering step of estimating a group having similar characteristics from a data set and clustering, the fuzzy membership function for the estimated cluster Calculating fuzzy membership using the program; and generating fuzzy rules for the input data with reference to the calculated fuzzy membership, wherein the clustering step uses Chiu's method for cluster estimation and potential potential of the cluster. Characterized in that it is used for clusters that estimate the center. And the fuzzy rule generation step is characterized by generating a fuzzy rule using a sugeno fuzzy inference model.

In the present invention, a method of predicting red tide to occur in the future using fuzzy inference and 10 days of temperature / temperature / precipitation information has been proposed. The proposed method was coclodinium p. P. We learned using red tide information and evaluated the proposed method using information from 2008 to 2010. The evaluation showed that the average forecast accuracy improved over three years.

1 is a block diagram for explaining red tide occurrence prediction according to an embodiment of the present invention.
2 is a fuzzy membership of a cluster of learning materials.
3 is a comparison result of accuracy rates for red tide occurrence prediction using the BPNN method, GRNN method, SVM method and fuzzy inference method according to an embodiment of the present invention.

BRIEF DESCRIPTION OF THE DRAWINGS The foregoing and further aspects of the present invention will become more apparent from the following detailed description of preferred embodiments with reference to the accompanying drawings. Hereinafter, the present invention will be described in detail to enable those skilled in the art to easily understand and reproduce the present invention.

1 is a block diagram illustrating a red tide occurrence prediction according to an embodiment of the present invention.

In the preprocessing 100 of FIG. 1, the learning material is processed to predict the red tide occurrence. The study data were taken from the occurrence data of coclodinium p., A red tide organism in the Tongyeong region of 2002-2007, and the water temperature, temperature, and precipitation of the region. Among them, the red tide occurrence information was collected from the National Institute of Fisheries Science, the red tide information system, and the water temperature information was collected from the coastal stop observation information of the Ocean Fisheries Research Information Portal. Table 1 shows the density of red tide biomass in 2005 when fewer red tide were found in the study data. And Table 2 is the water temperature, temperature, and precipitation information for 10 days before the red tide of Table 1, Table 3 shows the number of red tide occurrence, the number of days of temperature and temperature before the red tide occurs, the number of precipitation.

As pretreatment for the study data, the water temperature and temperature calculate the average value for 10 days before the red tide occurs, and the precipitation calculates the total precipitation for the 10 days before the red tide occurs. In addition, the biodensity is divided by 1,000 and normalized to calculate the red tide rate. Among them, 10-day average water temperature, 10-day average temperature, and 10-day total precipitation are input learning data and red tide rate is output learning data. The 10-day water temperature average, 10-day temperature average, and 10-day total amount of precipitation are the results of pretreatment for the study data.

As shown in FIG. 1, the fuzzy inference 200 is divided into a cluster 2100, a fuzzy membership calculation 220, and a fuzzy rule generation 230. In order to predict the occurrence of red tide using fuzzy inference, a rule is required to determine whether the input data is the occurrence of red tide. To do this, it is necessary to cluster learning materials into groups with similar characteristics, and generate rules by calculating fuzzy membership of the clustered materials.

In order to generate fuzzy rules, it is necessary to classify the learning materials into groups for each characteristic. To this end, according to an aspect of the present invention, a subtractive clustering method used in fuzzy inference in the cluster 210 process is used. The subtractive clustering method is a method of estimating a group having similar characteristics from a data set. In one embodiment, the fuzzy inference 200 uses Chiu's method for cluster estimation, and estimates the potential center of the cluster using Equation 1 for clustering.

Where P _i is the center of the i th potential cluster, r is the positive coefficient defined by the surrounding radius, x _i is the potential data belonging to the i th cluster, and x _j is the j th training material.

In the fuzzy membership calculation 220, fuzzy membership is calculated using a fuzzy membership function of one of a trigangluar, a gaussian, and a trapezoid. Preferably, in the fuzzy membership calculation process 220, a Gaussian function is used to express detailed membership. The Gaussian membership function may be calculated as shown in Equation 2, and the calculated membership represents a broad curve as shown in FIG.

Where x is the learning material, σ is a constant that determines the width of the Gaussian curve, and c is a constant that determines the center of the Gaussian curve. In one embodiment, the initial values were set so that the width of the Gaussian curve was 2 and the center of the curve was centered using σ = 2, c = 5.

In an embodiment, the fuzzy rule generation 230 uses a sugeno fuzzy model that is frequently used in fuzzy inference. Table 4 shows fuzzy rules generated using the sugeno fuzzy model.

The red tide rate, which is the final output value of the output rule of the sugeno fuzzy model, is calculated as in Equation 3.

Where w is the weight for fuzzy rules, z is the cluster for red tide membership, and N is the number of rules. w _i = f (10-day average temperature cluster) and f (10-day average temperature cluster) and f (10-day total precipitation cluster), where wi is the weight of the i th rule and f () is the Gaussian membership of Equation 2. Function.

FIG. 2 is a diagram illustrating a membership value calculated using Equation 2, showing the results of grouping learning data for water temperature / temperature / precipitation / red tide rate using Equation 1 into six groups. In FIG. 2, when the 10-day average water temperature is 25.6, the 10-day average temperature is 29.1, and the 10-day total precipitation is 241 as an input value, the red tide rate is calculated to be 3.74 by the fuzzy rule and Equation 3 of Table 4. . That is, red tide generation is predicted the next day. If the red tide rate is 0 or less, it is predicted that red tide will not occur.

3 is a diagram illustrating red tide occurrence prediction using a back propagation neural network (BPNN) method, a general regression neural network (GRNN) method, a support vector machine (SVM) method, and a fuzzy inference (FR) method according to an embodiment of the present invention. Figure of accuracy comparison results. It is confirmed from FIG. 3 that the accuracy factor of the fuzzy inference method according to the present invention is the highest.

So far I looked at the center of the preferred embodiment for the present invention. It will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the spirit and scope of the invention as defined by the appended claims. Therefore, the disclosed embodiments should be considered in an illustrative rather than a restrictive sense. The scope of the present invention is defined by the appended claims rather than by the foregoing description, and all differences within the scope of equivalents thereof should be construed as being included in the present invention.

100 pretreatment 200 fuzzy inference
210: Cluster 220: Fuzzy membership calculation
230: Create fuzzy rules

Claims

A clustering step of estimating and clustering a group having similar characteristics from the data set;
Calculating fuzzy membership for the estimated cluster using a fuzzy membership function; And
Generating a fuzzy rule for the input data with reference to the calculated fuzzy membership;
The clustering step uses Chiu's method for cluster estimation and uses the cluster for estimating the potential center of the cluster.

The method of claim 1,
The fuzzy rule generation step includes generating a fuzzy rule using a sugeno fuzzy inference model.