Automatic Group Decision-Making for Algal Bloom Management Based on Information Self-Learning
[ad_1]
1. Introduction
All of these studies adopted the method of group decision-making. Group decision-making can integrate the diverse perspectives and experiences of different members to allow for a more comprehensive analysis that can help identify the best solution to a problem. It can reduce the subjective bias and blind spots of individuals, improve the objectivity and fairness of decision-making, and reduce the possibility of serious errors in decision-making. After the research, it is found that the problem of algal bloom outbreaks has been paid attention to all over the world, and researchers have conducted in-depth analyses of the causes of algal bloom outbreaks, and some of them have focused on the mechanism of algal bloom management methods, but most of the decision-making of water bloom management in most of the regions still relies on human decision-making, which is mainly based on experience and common sense, and has not gone through the rigorous analysis of data; therefore, this method makes it difficult to cope with the complexity of the problem or unknown problems, and is highly susceptible to the interference of subjective factors. The previous explorations of our team members mainly focused on introducing rational decision-making into the decision-making of algal bloom, and developed a detailed decision-making process, which laid a good foundation for the subsequent in-depth research. Although these studies have used multiple-attribute decision-making, they have not conducted in-depth research on the weighting problem in the decision-making process; most of the weights are given by human beings, and the calculation process is relatively vague, which does not take advantage of the objective data statistics and analyses.
In the research on multi-attribute decision-making methods, most scholars focus on the proposal and optimization of decision-making structural models and introduce fuzzy theory into decision-making in a way that is more flexible and closer to human cognition. Therefore, it is very reliable to introduce the multi-attribute decision-making method into the decision-making of algal bloom. Aiming at the characteristics of algal bloom decision-making with many management methods and rich management objectives, this study invites several water environment experts to make group decision-making, and their decision-making opinions are shown in a richer and more detailed expression—a three-parameter interval number. A density-clustering algorithm is introduced in multi-attribute decision-making, using its information self-learning characteristics to automatically obtain the weight information of different experts and methods, so that the decision-making information of each expert plays a different role. Finally, it obtains the comprehensive ranking results of the methods under the influence of multi-objectives, decides on the best management method, and realizes scientific automatic rational decision-making.
Aiming at the above difficulties in decision-making, we introduce the multi-expert-group processes method in decision-making, The use of group processes allows for the synthesis of various views and experiences and determine the weights more objectively. Information self-learning algorithms are integrated into multi-objective decision-making to automatically select from a certain number of alternatives that can fully meet the given goal. On the issue of fuzziness, the evaluation index is transformed into the three-parameter interval number, which can provide a more accurate and comprehensive result, and optimize the overall decision-making algorithm by combining the characteristics of the three-parameter interval number, so that the decision-making of the algal bloom management can be made more accurately and scientifically. Finally, we describe the proposed method to study the Sanjiadian Reservoir in Beijing.
2. Materials and Methods
2.1. Decision-Making Tool Structure
For the sustainable management of algal bloom in urban lakes, a suitable and optimal management method should be selected first. The optimal choice of the method is the decision problem studied in this paper.
First of all, the water environment monitoring technology and system is used to monitor the water quality, obtain and analyze the parameters that characterize the status of algal bloom, including water quality, hydrological data, and meteorological conditions, and promptly classify the monitoring results according to the standard of algal bloom classification based on the level, if the status of algal bloom is normal, then continue to monitor it, and if the level of the status of algal bloom has reached the alert value, then immediately enter the emergency state, and analyze the decision-making through the system to obtain the best method for the emergency situation. The inputs to the decision-making model and algorithm include quantitative water quality parameters, and the subjective opinions of experts and management objectives, which can result in the decision values under the influence of the decision-making objectives, and ultimately provide the results of the advantages and disadvantages of the management programs to the decision-makers of the environmental protection department.
where k is the sequence number of the expert and the matrix element indicates the degree of satisfaction of the i-th method to the j-th objective. The experts give the degree of satisfaction of each method for the different objectives in the rating scale. The form of evaluation in this paper is expressed in terms of the number of a three-parameter interval number, which covers the information more comprehensively and allows the experts to give a more fine-grained description of the evaluation of the methods.
The main tasks of the paper are two. First, integrate the experts’ evaluations of each method, and in the process of integration, use the information self-learning ability of the improved the density-based spatial clustering of applications with noise (DBSCAN) to obtain the weights automatically. Secondly, the results of multi-objective, multi-expert cluster decision-making are obtained, and the integrated evaluation values of each method are ranked using the likelihood ranking method based on the three-parameter interval number to obtain the final decision-making results.
2.2. Study Area and Experimental Data
This paper analyzes the Sanjiadian Reservoir in Mentougou District, Beijing as an example, proving that this decision-making method can be clearly and accurately applied in the management of Sanjiadian waters, and that the method can be applied in other similar management.
Beijing’s Sanjiadian Reservoir, located west of Sanjiadian Village in Mentougou District, is a large-scale water conservancy hub project on the main stream of Yongding River.
Sanjiadian Reservoir stops the Yongding River, so that the rushing river water can rest for a long time and achieve precipitation purification here. At present, the water quality of Sanjiadian Reservoir reaches the standard of three types of surface water, which basically meets the water quality requirements of drinking water sources, and is ready to activate the Guanting Reservoir at any time to supply water to the urban area in case of emergency.
2.3. Decision-Making Framework
In recent years, the research on decision-making theory and methods has developed in the direction of information and intelligence. In the era of information intelligence, scientific decision-making should be done in such a way as to take full account of the empirical knowledge of managers and experts, water pollution control objectives, and the real-time environment of lakes and reservoirs, which involve a wide range of sources and types of information, and the decision-making mechanism of algal bloom control needs to solve the problem of comprehensive processing and analysis of diversified information.
The initiation of the decision-making process for algal bloom management is determined by the bloom outbreak, and algal bloom in rivers and lakes generally pass through the germination stage, growth stage, outbreak stage, and extinction stage. Based on the analysis and judgement of the water quality monitoring information, the growth status of the bloom can be determined, and when the main phenological indicators of the bloom exceed the preset thresholds, the treatment decision-making method is initiated.
The decision-making process has three main components. The yellow block is the real-time monitoring of water quality status. The processing of expert opinions and the calculation of target weights is shown in the green blocks, and the blue block is the calculation of target weights and the integration and ranking of decision-making opinions.
2.4. Conversion Method for Three-Parameter Interval Number
In the group decision-making process, due to the complexity and uncertainty of the decision problem and the differences in the personal preferences of the decision makers, it is often impractical to use a deterministic form of preference to portray a complex problem, and mathematical methods of uncertainty, such as interval numbers, fuzzy numbers, random variables, etc., have been widely used in the field of decision-making.
The source of information for decision-making on algal bloom management is the decision-making evaluation opinions of experts, and it is set up to invite a number of experts in the field to publish their decision-making opinions on the same lake and reservoir algal bloom management problem, the expert set is
. In order to facilitate computer processing, the form of the statements in which the experts expressed their opinions was restricted, and all of them were in the form of a three-parameter interval number. A three-parameter interval number means that three parameters are used to represent an interval number, for example
and , where, and are the upper and lower limits of the interval, is the most probable number in the interval, called the center of gravity of the interval.
The method covers the information more comprehensively and enables the experts to describe the evaluation of the method more finely. Compared with the traditional interval number representation, the three-parameter interval number is simple to compute and can be easily used for basic operations such as addition, subtraction, multiplication, and division, and is also adapted to fields such as fuzzy mathematics and grey theory. When there are outliers or errors in the data, this interfering information can be better eliminated, providing a more robust description of the data. Three-parameter interval numbers are widely used in practical applications and have many advantages.
For three-parameter interval numbers, there are the following algebraic algorithms as shown in Equations (2)–(5).
There are two three-parameter interval numbers,
.
2.5. A Multiple-Attribute Decision-Making Approach Based on Self-Information
Multiple-attribute decision-making is the process of integrating and ranking the criterion values of multiple solutions under multiple criteria. In decision-making for a particular problem, the goal that is usually set after the problem is solved, so it is hoped that the final selected method can satisfy the given goal, and the method of multiple-attribute decision-making is to select the method that can comprehensively satisfy the given goal from a certain number of alternatives through a certain method.
In the figure, A is the core point, the number of points in the radius r around it is greater than or equal to the minimum number of points (minPts), then these points will be classified as a cluster. B and C represent the border points, the number of points in the radius r around it is less than the minPts, but it belongs to the boundary of a certain cluster and will be classified into this cluster. N is the noise point, the radius r around it is also less than the minPts, is regarded as outliers or noise points, and does not belong to any cluster. N is an outlier, the number of points in radius r around it is also less than the minPts, it is regarded as an outlier, and it does not belong to any cluster. DBSCAN algorithm does not need to pre-specify the number of clusters and the shape of the clusters, but it is automatically determined by the density threshold, so for some complex datasets, DBSCAN can achieve better results than the traditional clustering algorithm.
The original DBSCAN algorithm uses the formula for calculating the Euclidean distance, and the data in this study is in the form of a three-parameter interval number; in order to take advantage of this interval number, we improve the algorithm of DBSCAN. When calculating the distance, different weights are given to the upper and lower limits and the center of gravity, and it is guaranteed that the center-of-gravity point has the largest weight; this distance formula can be adjusted according to the decision-maker’s experience and risk appetite, and the improved distance algorithm is shown in Equation (6).
The multi-objective decision-making steps based on self-learning of information are given according to the algorithm of the three-parameter interval number.
The comprehensive judgement data is used as input for the density-based clustering algorithm for clustering, and the weights are automatically obtained based on the clustering results.
The training algorithm is as follows (Algorithm 1):
Algorithm 1: Self-learning algorithm for expert evaluation information |
Input: DB: Database, ɛ:Radius, minPts: Density threshold, dist: Distance function. Data: label: Point labels, initially undefined. Output: (Weighted data, cluster). foreach point p in database DB do if label(p) ≠ undefined then continue Neighbors N ← RangeQuery(DB, dist, p, ɛ) If |N| < minPts then Label(p) ← Noise noise number + 1 noisy+ = p continue c ← next cluster label label(p) ← c Seed set S ← N\{p} foreach q in S do if label(q) = Noise then label(q) ← c if label(q) ≠ undefined then continue Neighbors N ← RangeQuery(DB, dist, q, ɛ) Label(q) ← c If |N| < minPts then continue S ← S ∪ N number + 1 all+ = q r = all/q end for return r, number end for R = (r × number) + (noisy × noise number) Return R |
The evaluation information of each method is automatically clustered and the weights are obtained to achieve the self-learning of the information of the method agglomeration, and, finally, the comprehensive evaluation value of R each method by k experts can be obtained.
i.e., Number of two three-parameter intervals:
,
and
,
,
.
3. Results and Discussion
According to the decision-making mechanism for algal bloom management, the decision-making process starts when the real-time data reaches the set threshold. In accordance with the decision-making management framework, experts are first asked to provide an evaluation of the set options based on the current lake status and to rank the management options according to the management needs of the area. In this section, the experimental results are presented using Sanjiadian Reservoir as an example.
The water of the Sanjiadian Reservoir on the Yongding River flows into Beijing from the Youzhou Gorge in Guanting Town, Huailai, Hebei Province. After investigation, most of the algae in this water area are harmful cyanobacteria such as Microcystis. In the study of Sanjiadian Reservoir, the decision-making target set {overall input scale, degree of bloom clean-up, degree of environmental impact, treatment efficiency, and degree of sustainability}, the alternative set {salvage machinery, air flotation, ultrasonic method, adsorption method, algaecide, coagulating sedimentation, electrochemical, algophagous, and Microorganisms}, and the seven experts were selected from a large number of experts, and the opinions were obtained in the form of a moderated questionnaire.
According to the decision-making steps, first we obtain the objective weight set, according to the location and function of Sanjadian Reservoir, the weight set of the five management objectives is = {(0.4,0.5,0.55), (0.3,0.44,0.56), (0.45,0.5,0.6), (0.2,0.3,0.35), (0.1,0.25,0.3)}, perform normalization calculations is = {(0.2,0.25,0.27), (0.15,0.22,0.28), (0.22,0.25,0.3), (0.2,0.15,0.17), (0.05,0.12,0.15)}.
From P, the order of advantages and disadvantages of the 9 options is as follows A1, A8, A2, A9, A3, A5, A4, A6, A7. i.e., {salvage machinery > algophagous > air flotation > microorganisms > adsorption method > algaecide > ultrasonic method > coagulating sedimentation> electrochemical.
The analysis of the decision-making results in Sanjiadian Reservoir reveals that the managers’ objectives revolve around two main aspects: the degree of environmental impact and the overall investment scale. These two factors play a crucial role in determining the prioritization of treatment methods. In line with the characteristics identified during the early research phase, the mechanical algal-removal method and the algal-feeding biological method emerge as the top-ranked options. The high ranking of these methods can be attributed to their favorable attributes in terms of environmental impact and investment requirements. The decision-makers aim to strike a balance between achieving effective algal bloom control and managing costs efficiently. The findings of this analysis provide valuable scientific references for the managers involved in algal bloom treatment decision-making. By considering a range of factors, including environmental impact, investment requirements, and treatment efficiency, the decision-makers can make informed choices that best meet their objectives and the specific needs of the Sanjiadian Reservoir. This enhanced decision-making process ensures a more balanced and sustainable approach to algal bloom management, ultimately benefiting both the ecosystem and the management goals.
The methodology of this paper has the following advantages over the previous related studies conducted by our team members. First, the evaluation data are expressed in a more detailed form, and the three-parameter interval numbers express richer and more comprehensive information than the traditional intervals, which is very important for decision-making relying on the evaluation data. Second, in terms of the selection and optimization of clustering methods, this method combines the density clustering algorithm with the three-parameters interval number. The density clustering algorithm does not need to pre-specify the number of clusters, and it is capable of automatically detecting and determining the number of clusters in the data, which reduces the need for subjective interventions and provides a more flexible clustering analysis. While expert evaluation data varies and is not centralized, density clustering algorithms work well with datasets that have irregular shapes, noise, and outliers. It makes all the experts’ evaluation data available for the decision-making process, which makes the decision-making results more comprehensive and scientific.
Automatic group decision-making methods can streamline the decision-making process by quickly gathering input from multiple stakeholders and generating consensus-based decisions. The algorithm in this study can help improve the accuracy of decision-making regarding algal bloom management. This can lead to more effective strategies for mitigating the impact of algal blooms on ecosystems and ensuring water safety. This method can adapt quickly to changing conditions and new information, allowing decision-makers to adjust strategies in real-time as new data on algal blooms and water quality become available. This method can lead to faster responses to algal bloom events and water safety issues.
4. Conclusions
Algal bloom management decision-making is a complex system engineering problem. Based on the analysis of the current situation of algal bloom management decision-making, this study proposes a group decision-making method based on self-learning of information, integration, and sorting of experts’ opinions with regard to the subjectivity caused by human-assigned weights in the decision-making process and the inability to use the advantages of data. The automatic rational decision-making process of multi-objective, multi-expert, and self-learning of decision-making information is realized. It also applies the three-parameter interval number to the decision-making process, and optimizes the original algorithm according to its characteristics, so that the experts’ opinions can be applied more directly and accurately in the decision-making process. The experiments in Beijing Sanjiadian Reservoir also show that this method can help managers to choose the management method more scientifically according to the actual engineering needs.
We provide a more detailed discussion of the applicability of the algal bloom management decision-making based on information self-learning elsewhere. The data sources for this decision-making method are mainly water quality monitoring stations and evaluation indicators from experts in the field. These data are finally transformed into a set of intervals for decision analysis. Therefore, if the decision-making data is a set of intervals, the method can be used for decision-making.
However, this paper has only conducted an in-depth study of decision data fusion techniques in multi-attribute decision-making, and has not considered the overall decision-making process or other forms of decision-making information; therefore, more information can be introduced into decision-making, such as water quality prediction data, in future work. Subsequently, multi-source data, including satellite remote-sensing data, water quality monitoring data, and ecosystem parameters, can be combined to provide more comprehensive and multi-dimensional information support for group decision-making, so as to enhance the comprehensiveness and reliability of decision-making. This method still requires expert intervention in data processing methods; therefore, the investment in manpower and time is still too high. Subsequent research could delve deeper into data transformation and representation in order to automate the overall decision-making process. At the same time, further research can also be carried out on decision-making information-mining work to achieve more scientific and faster intelligent decision-making.
[ad_2]