CN112785004A - Greenhouse intelligent decision-making method based on rough set theory and D-S evidence theory - Google Patents
Greenhouse intelligent decision-making method based on rough set theory and D-S evidence theory Download PDFInfo
- Publication number
- CN112785004A CN112785004A CN202110150119.7A CN202110150119A CN112785004A CN 112785004 A CN112785004 A CN 112785004A CN 202110150119 A CN202110150119 A CN 202110150119A CN 112785004 A CN112785004 A CN 112785004A
- Authority
- CN
- China
- Prior art keywords
- greenhouse
- decision
- theory
- rough set
- evidence
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 33
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 48
- 230000009467 reduction Effects 0.000 claims abstract description 23
- 238000005315 distribution function Methods 0.000 claims abstract description 13
- 238000012545 processing Methods 0.000 claims abstract description 13
- 238000004364 calculation method Methods 0.000 claims abstract description 11
- 239000011159 matrix material Substances 0.000 claims abstract description 11
- 238000009826 distribution Methods 0.000 claims description 18
- CURLTUGMZLYLDI-UHFFFAOYSA-N Carbon dioxide Chemical compound O=C=O CURLTUGMZLYLDI-UHFFFAOYSA-N 0.000 claims description 16
- 239000002689 soil Substances 0.000 claims description 10
- 229910002092 carbon dioxide Inorganic materials 0.000 claims description 8
- 239000001569 carbon dioxide Substances 0.000 claims description 8
- 238000005286 illumination Methods 0.000 claims description 7
- 230000002776 aggregation Effects 0.000 claims description 4
- 238000004220 aggregation Methods 0.000 claims description 4
- 238000010606 normalization Methods 0.000 claims description 3
- 238000005096 rolling process Methods 0.000 claims description 3
- 239000013598 vector Substances 0.000 claims description 3
- 238000010801 machine learning Methods 0.000 abstract description 6
- 238000012795 verification Methods 0.000 abstract 1
- 230000006870 function Effects 0.000 description 16
- 230000004927 fusion Effects 0.000 description 9
- 238000012360 testing method Methods 0.000 description 6
- 238000012549 training Methods 0.000 description 6
- 230000000694 effects Effects 0.000 description 4
- 238000012544 monitoring process Methods 0.000 description 4
- 230000009471 action Effects 0.000 description 3
- 238000011161 development Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 230000005540 biological transmission Effects 0.000 description 2
- 230000015572 biosynthetic process Effects 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 150000001875 compounds Chemical class 0.000 description 1
- 238000002790 cross-validation Methods 0.000 description 1
- 238000013144 data compression Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000012067 mathematical method Methods 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 238000003786 synthesis reaction Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
- G06N20/10—Machine learning using kernel methods, e.g. support vector machines [SVM]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/02—Agriculture; Fishing; Forestry; Mining
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Software Systems (AREA)
- Business, Economics & Management (AREA)
- Marine Sciences & Fisheries (AREA)
- Primary Health Care (AREA)
- Mining & Mineral Resources (AREA)
- Agronomy & Crop Science (AREA)
- Health & Medical Sciences (AREA)
- Economics (AREA)
- General Health & Medical Sciences (AREA)
- Human Resources & Organizations (AREA)
- Marketing (AREA)
- Animal Husbandry (AREA)
- Strategic Management (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Medical Informatics (AREA)
- Probability & Statistics with Applications (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a greenhouse intelligent decision-making method based on a rough set theory and a D-S evidence theory, which relates to the technical field of intelligent agriculture, wherein processed data are obtained through fuzzy C-means clustering processing and attribute reduction of a kernel and a rough set; and constructing a basic probability distribution function by using the rough set, and calculating the support degree among all the influence factors of the greenhouse. And (3) introducing the BPA basic probability assignment matrix obtained by calculation by using an improved D-S evidence theory, and constructing a confidence matrix to complete the combination of greenhouse influence factors to obtain a decision result. A BPA basic probability assignment matrix is introduced by utilizing a traditional SVM algorithm aiming at small sample machine learning to obtain a decision result, and algorithm comparison verification is carried out with a D-S evidence theory algorithm.
Description
Technical Field
The invention relates to the technical field of intelligent agriculture, in particular to a greenhouse intelligent decision method based on a rough set theory and a D-S evidence theory.
Background
Agriculture is an important national pillar industry, and the promotion of greenhouse influence factor intelligent decision technology is the most important factor in the agricultural development of China. Intelligent decision making technology for greenhouse influencing factors is mainly used for adjusting the external conditions of crop growth. Each influencing factor is different from crop to crop, and it is necessary to provide the correct influencing factors to adjust and ensure good growth of the crops.
Due to the influence of various unstable factors such as the change of influence factors in the greenhouse and the internal hardware problem of the monitoring equipment, the original data acquired by the monitoring equipment is not necessarily correct, and for the data which are not necessarily correct, if the data are processed by an intelligent decision algorithm, the situation of judgment error or decision error is likely to occur.
One of the main features of the D-S evidence theory is that in the presence of uncertain data, the data can be processed efficiently. Most of data collected by the detection equipment are continuous original data, and the processing capacity of the computer on the continuous data is poor, so that after the continuous data are discretized, a rough set theory is adopted, and the main reason for adopting the rough set theory is that the characteristic of attribute reduction in the rough set theory can remove useless condition attributes, and the original function of the data is unchanged, so that the operation cost of the algorithm can be effectively reduced.
In the aspect of monitoring greenhouse influence factors, most of the monitoring transmission of the greenhouse carries out data transmission by communication technologies such as Wi-Fi and ZigBee. The greenhouse environment is a system with large inertia and nonlinear characteristics, and has the phenomena of time delay, cross connection and the like, so that the difficulty is very high when an accurate model is required to be established.
Aiming at the problems in the prior art, the application provides a greenhouse intelligent decision method based on a rough set theory and a D-S evidence theory, the improved D-S evidence theory is applied, a BPA basic probability assignment matrix obtained through calculation is introduced, a confidence matrix is constructed to complete the combination of greenhouse influence factors, and a greenhouse intelligent decision result is obtained.
Disclosure of Invention
The invention aims to provide a greenhouse intelligent decision method based on a rough set theory and a D-S evidence theory.
The invention provides a greenhouse intelligent decision method based on a rough set theory and a D-S evidence theory, which comprises the following steps:
step 1: collecting greenhouse influence factor sample data, establishing a rough set and a D-S evidence theoretical model, and carrying out attribute reduction on the greenhouse influence factor sample data by taking the rough set theory as the front edge of the model;
step 2: obtaining basic confidence coefficient distribution of a combination result by using a D-S evidence combination rule and combining greenhouse influence factor sample data;
and step 3: discrete processing is carried out on the sample data by using a fuzzy C-means clustering method, and n vectors x are processedi(i-1, 2, …, n) into c fuzzy groups and finding the cluster center for each fuzzy group, minimizing the value function of the dissimilarity index, using [0,1 ] for each given data point]The degree of membership between each group is determined, and according to the normalization, the sum of the degrees of membership of one dataset is always equal to 1:
the value function of the fuzzy C-means clustering algorithm is as follows:
wherein u isijIs at [0,1 ]]C isiFor the cluster center of I in the fuzzy C-means cluster, dij＝||ci-xj| | is the euclidean distance between the T-th clustering center and the j-th data point, and m > -1 is a weighting index;
to gatherClass center CiIs calculated as:
the membership degree is calculated as:
and 4, step 4: when the corresponding clustering category of each sample is determined, determining to use a threshold value with overlapped membership degree, and selecting all sample data corresponding to aggregation for limitation by using the threshold value of the sample data value;
and 5: constructing a basic probability distribution function by using the rough set, and calculating the support degree among all influence factors of the greenhouse;
step 6: and (3) introducing the BPA basic probability assignment matrix obtained by calculation by using an improved D-S evidence theory, and constructing a confidence matrix to complete the combination of greenhouse influence factors to obtain a decision result.
Further, the sample data of greenhouse influencing factors collected in the step S1 includes temperature, humidity, illumination intensity, temperature in soil, humidity in soil and volume fraction of carbon dioxide.
Further, the decision result mentioned in the step S6 includes four categories 1, 2, 3, and 4, which correspond to the operations of opening the roll film, opening the roll film and starting the fan, starting the fan and the wet curtain, and not acting, respectively.
Further, the decision result is set as { l (k) }, k is 1, 2, 3, 4}, where k is the result of 4 decisions, the basic credibility distribution function in the power set of the greenhouse influence factor control decision identification framework represents the support degree of the greenhouse influence factor information of various decision categories, and
∑A∈Θm (A) is 1, wherein m (1), m (2), m (3) and m (4) represent basic credibility distribution of greenhouse decision factors, and m (phi) representsBasic credibility allocation of uncertain greenhouse influencing factors.
Further, the set of temperature values corresponding to the basic reliability distributions M (1), M (2), M (3), and M (4) of the greenhouse decision factors is {31.5 ℃, 32.8 ℃, 29.2 ℃ }, {35.1 ℃, 36.0 ℃, 34.7 ℃ }, {38.2 ℃, 42.3 ℃, 37.5 }, {24.9 ℃, 27.0 ℃, 25.1 }.
Compared with the prior art, the invention has the following remarkable advantages:
the invention provides a greenhouse intelligent decision method based on a rough set theory and a D-S evidence theory, which firstly researches the processing of a fuzzy C-means algorithm (FCM) on data, carries out discretization processing on continuous data so as to carry out attribute reduction and decision rule reduction of the rough set theory on original data, secondly screens out unnecessary condition attributes in expert knowledge by using an attribute reduction method based on information entropy, and finally carries out decision on data by using D-S evidence fusion, determines decision categories and selects a threshold value by using a decision based on basic credibility distribution: epsilon1＝0.2，ε2And (5) obtaining the D-S evidence combination and decision to obtain a decision result. Meanwhile, the SVM algorithm learned by the machine of the small sample is used for processing the same group of data to obtain a decision result of the SVM algorithm, and the results of the two algorithms are compared and verified. The operation cost of the decision method based on the rough set and the D-S evidence theory is far less than that of the SVM algorithm of small sample machine learning, the calculation complexity is low, and the processing effect is better.
Drawings
FIG. 1 is a diagram of a decision model based on rough set and evidence theory provided by an embodiment of the present invention;
FIG. 2 is a comparison of the basic confidence levels generated after the combination of evidence through D-S evidence theory.
Detailed Description
The technical solutions of the embodiments of the present invention are clearly and completely described below with reference to the drawings in the present invention, and it is obvious that the described embodiments are some embodiments of the present invention, but not all embodiments. All other embodiments, which can be obtained by a person skilled in the art without any inventive step based on the embodiments of the present invention, shall fall within the scope of protection of the present invention.
The Rough Set Theory was created in 1982 by the scientist z, pawlak in poland, and has resulted in significant development in the field of data processing and analysis.
Attribute reduction and rule reduction are important research topics in rough set theory. Since knowledge in the knowledge base is not every one useful and not every one important, on the contrary, there are some redundant knowledge which occupies resources, which causes resource loss and also interferes people or computers to make correct judgment. The attribute reduction of the data is to ensure that the decision classification capability of the original data is unchanged and to delete unnecessary condition attributes in the original data. The process of reducing the dependency between the decision attribute and the condition attribute does not change.
Definitions 1-I ═ (C ═ D), C, D are the conditional and decision attributes, respectively, C ∈ C, if γ is equal to CC(D)＝γC-{c}(D) C is reducible in C, independent if any C e C is not melodic in C, otherwise C is relevant. All irreducible relations in C are called sum-kernels, and the set formed by the sum-kernels is called a kernel set and is marked as CORE (C).
in the formula (I), the compound is shown in the specification,
the shafer introduces the concept of a trust function into a D-S evidence theory, which is a great improvement on the D-S evidence theory, thereby generating a set of scientific mathematical methods capable of processing multi-data fusion. The Bayesian inference method is the basis of the D-S evidence theory, and the Bayesian conditional probability is the key for realizing the Bayesian inference method. The concept of Basic Probability Assignment (BPA) is the confidence that the framework assigns to each proposition, and m (A) is the basic confidence, reflecting the reliability of A. Let two evidence sources of the basic probability distribution function be m1 and m2, and the calculation of the basic probability distribution function is derived from the existing D-S evidence theory combination rule, which is the basic process of D-S evidence fusion.
Assuming U as the recognition frame, the function m: 2u→[0，1]The following conditions are satisfied:
(1)
(2)
the basic assignment of a is called m (a) ═ 0, and m (a) ═ 0 denotes the degree of trust in a, also called the mass function.
Bel：2u→[0，1] (6)
Represents the sum of the probability distribution functions of all subsets of a.
Defining a composition of 3 more trust functions
Let Bel1，…，BelnIs a trust function on the same identification framework theta, and the basic credibility of each proposition on the framework is assigned as m1, … and mn, if the basic credibility is m1, … and mn
Meaningful, the basic credibility distribution after the theoretical synthesis of the D-S evidence can be calculated as m, and then
A1，…，AnE theta can be calculated by equation (8) and can be basically calculatedAnd (4) assigning a value of the confidence level.
In the D-S evidence theory, it is most important to determine the basic probability distribution function of the focus element, which is also a key point of whether the D-S evidence theory can be implemented in practical application. Meanwhile, many aspects such as evidence acquisition and basic credibility distribution also depend on the experience of field experts or historical data, and subjectivity is difficult to avoid. However, due to the characteristics of the rough set theory, the concept of membership can be adopted, that is, the rough set theory is improved to be capable of analyzing and identifying a frame by analyzing the membership of an object, so that the purpose of acquiring basic credibility distribution can be achieved. In conclusion, the rough set theory is suitable for being combined with the D-S evidence theory, helps the D-S evidence theory to solve the important problem that the evidence is difficult to obtain, and can also solve the fault caused by the expert experience judgment.
Referring to fig. 1-2, the invention provides a greenhouse intelligent decision method based on rough set theory and D-S evidence theory, comprising the following steps:
step 1: collecting greenhouse influence factor sample data, wherein the greenhouse influence factor sample data comprises temperature, humidity, illumination intensity, soil temperature, soil humidity and carbon dioxide volume fraction; establishing a rough set and a D-S evidence theoretical model, and carrying out attribute reduction on greenhouse influence factor sample data by taking the rough set theory as the front edge of the model;
step 2: obtaining basic confidence coefficient distribution of a combination result by using a D-S evidence combination rule and combining greenhouse influence factor sample data;
and step 3: discrete processing is carried out on the sample data by using a fuzzy C-means clustering method, and n vectors x are processedi(i-1, 2, …, n) into c fuzzy groups and finding the cluster center for each fuzzy group, minimizing the value function of the dissimilarity index, using [0,1 ] for each given data point]Degree of membership between each group, according to a normalization rule, a datumThe sum of membership of the sets is always equal to 1:
the value function of the fuzzy C-means clustering algorithm is as follows:
wherein u isijIs at [0,1 ]]C isiFor the cluster center of I in the fuzzy C-means cluster, dij＝||ci-xjI is the Euclidean distance between the I-th clustering center and the j-th data point, and m>1 is a weighted index;
and cluster center ciIs calculated as:
the membership degree is calculated as:
the fuzzy C-means clustering method is used for discrete processing of continuous data, and means that all data have the highest similarity in the same cluster, and the data have different attributes in different clusters and are opposite. The fuzzy C-means clustering algorithm is the development and improvement of the common C-means clustering algorithm. The data division of the common C-means algorithm is difficult to distinguish, and the fuzzy C-means clustering algorithm is equivalent to easy distinguishing and has higher flexibility. The fuzzy C-means clustering algorithm can optimize the objective function more effectively, can classify the original data better, can distribute the membership degree of the original data in the sample with efficiency, and provides a foundation for the automatic formation of the original data. After discretization is carried out on continuous data through a fuzzy C-means clustering algorithm, the category of the maximum membership degree of a sample of each discretized original data is used for recording, the category is acted on the sample value of the discretized data, the discretization result of the continuous data is used as a discretization set so as to be more effectively analyzed, and then a rough set theory is used for knowledge mining;
and 4, step 4: when the corresponding clustering category of each sample is determined, determining to use a threshold value with overlapped membership degree, and selecting all sample data corresponding to aggregation for limitation by using the threshold value of the sample data value;
wherein the membership function is a function representing the degree to which an object x is affiliated to the set A, commonly denoted as μA(x) The argument range is all possible objects belonging to the set A, and the value range is [0,1 ]]I.e. 0. ltoreq. muA(x)≤1。μA(x) 1 means that x completely belongs to the set A, which is equivalent to x ∈ A in the conventional set concept. A membership function defined in the space X ═ { X } defines a fuzzy set a, or fuzzy subset defined in the domain X ═ { X }. For a limited number of objects x1, x2, …, the xn fuzzy sets can be expressed as:
A＝{(μA(x)，Xi)|Xi∈X} (9)
and 5: constructing a basic probability distribution function by using the rough set, and calculating the support degree among all influence factors of the greenhouse;
step 6: and introducing a BPA basic probability assignment matrix obtained by calculation by using an improved D-S evidence theory, constructing a confidence matrix to complete the combination of greenhouse influence factors, and obtaining a decision result, wherein the decision result comprises four categories of 1, 2, 3 and 4, and the decision result respectively corresponds to the operation of opening a rolled film, opening the rolled film and starting a fan, starting the fan and a wet curtain and does not act.
Wherein, the decision result is set as { L (k) }, k is 1, 2, 3, 4}, wherein k is the result of 4 decisions, the basic credibility distribution function under the power set of the greenhouse influence factor control decision identification framework represents the support degree of the greenhouse influence factor information of various decision categories, and the basic credibility distribution function represents the support degree of the greenhouse influence factor information of various decision categories
∑A∈Θ m (a) 1, where m (1), m (2), m (3), m (4) represent the basic credibility distribution of greenhouse decision factors, and m (Φ) represents the basic credibility allocation of uncertain greenhouse influencing factors.
The basic reliability distributions of greenhouse decision factors, M (1), M (2), M (3), M (4), correspond to a set of temperature values {31.5 ℃, 32.8 ℃, 29.2 ℃ }, {35.1 ℃, 36.0 ℃, 34.7 ℃ }, {38.2 ℃, 42.3 ℃, 37.5 }, {24.9 ℃, 27.0 ℃, 25.1 }.
The data preprocessing method adopted by the invention is data reduction. Data reduction typically involves dimensionality reduction and data compression, meaning that researchers can use smaller conditional attributes, value ranges, and less data to express the same or similar information as the original data set. The rough set theory has the biggest characteristic of reducing condition attributes, and the condition attributes and the redundant attributes which can be removed are deleted through the knowledge of an expert table and the judgment of greenhouse influence factors. Rough set theory reduces the data set by compressing the data by replacing the equivalence class with objects in the equivalence class.
And (4) carrying out attribute reduction on the data of the greenhouse influence factor control expert knowledge table by taking the rough set theory as the front edge of the model on the rough set and the D-S evidence theory model. Then applying a D-S evidence combination rule in a D-S evidence theory and combining various greenhouse influence factors. These factors are combined with each other to obtain a basic confidence distribution of the combined result. This result is used to decide the action to take. A decision model based on rough set and evidence theory is shown in fig. 1.
After clustering by the fuzzy C-means algorithm, when determining the corresponding cluster class for each sample, it is decided to use the thresholds for membership overlap and to use the thresholds for sample data values to select all sample data for the corresponding aggregation for the constraint. For example, if the value of the data membership degree overlap of all sample data is greater than a threshold setting, and when the class selection of the sample data value setting threshold determines the class selection after the discretization of continuous data, it may be determined that the sample corresponds to a plurality of discrete classes. After the concept and judgment of the membership degree overlapping degree are added, the real situation can be reflected more truly. The degree of membership overlap introduced is practically equal to the limit range of membership values. For classification points belonging to a particular interval, a threshold of 0.4 may be used to determine the value of the membership value. The higher the threshold value, the clearer the boundary determined by the membership degree overlap. The more unique the corresponding class of sample data affected by noise, but this is not the case. On the contrary, the uniqueness is very weak, but the situation of the whole sample data can be reflected, and the situation is relatively true. Therefore, after the fuzzy C-means algorithm clustering, it is necessary to limit the discretized data by using the threshold of the degree of membership overlap.
Example 1
According to specific influence factor information of the greenhouse in the northwest region, an expert knowledge table for controlling the greenhouse influence factors is obtained, as shown in table 1, 12 groups of samples are counted, 6 greenhouse influence factors exist in each group of samples, a unique decision result is judged through the 6 influence factors, the greenhouse influence factors consist of 6 factors of temperature, humidity, illumination intensity, soil internal temperature, soil internal humidity and carbon dioxide volume fraction, the decision result comprises 4 types including 1 type, 2 type, 3 type and 4 type, namely, opening the rolling film and starting the fan, starting the fan and the wet curtain, and the fan and the wet curtain do not act. The data in table 1 contains greenhouse effect factor indicators and corresponding decision results, but the information contained in the greenhouse effect factor data is not easily understood by the user and therefore is difficult to directly use for decision making.
TABLE 1 greenhouse influence factor control expert knowledge table
And (3) clustering table 1 by using a fuzzy C-means clustering algorithm, wherein the clustering number is 4 and is equal to the number of the decision result categories, the clustering centers with algorithm return values are 4 grades, and the membership function of each sample belongs to 4 clustering centers respectively. See table 2, with no incompatible samples in table 2, indicating that the number of aggregates is suitable.
TABLE 2 greenhouse influence factor control decision-making table
And the table 2 is a decision table obtained after the FCM clustering algorithm, and is used for laying down an attribute reduction algorithm based on a rough set theory, so that the attribute reduction method is convenient to perform.
Example 2
According to an attribute reduction algorithm based on information entropy, a domain, temperature influence factor condition attributes and a decision result set are formulated, the attributes of a core are reduced when the temperature influence factor condition attributes are calculated relative to the core of the decision result, reduced data are obtained, and a cushion is laid for fusion of a D-S evidence theory.
In the decision of intelligent control of greenhouse influencing factors, influencing factors in the whole framework indicate which control method should be adopted. Thus, for table 1, the whole frame can be written as l (k), k ═ 1, 2, 3, 4, and k is the result of the 4 decisions. The basic credibility distribution function under the power set of the greenhouse influence factor control decision identification framework represents the support degree of the greenhouse influence factor information of various decision categories, and
∑A∈Θm (a) ═ 1, where m (1), m (2), m (3), m (4) represent the basic credibility distribution of greenhouse decision factors, and m (Φ) represents the basic credibility allocation of uncertain greenhouse influencing factors.
Let a, b, c be temperature, illumination intensity and carbon dioxide volume fraction, respectively. Taking temperature as an example, the basic credible interval is divided by the mean value of the environment indexes corresponding to each judgment category. The sets of temperature values corresponding to M (1), M (2), M (3), and M (4) are {31.5 ℃, 32.8 ℃, 29.2 ℃ }, {35.1 ℃, 36.0 ℃, 34.7 ℃ }, {38.2 ℃, 42.3 ℃, 37.5 ℃ }, {24.9 ℃, 27.0 ℃, 25.1 }, and the average values of the sets are 31.2 ℃, 35.3 ℃, 39.3 ℃, 25.3 ℃ respectively, and the basic credibility of the temperature factors is established as follows:
when a < 25.3 ℃, then m (4) ═ 0.9, m (1), m (2), m (3) are all 0, and m (Θ) ═ 0.1;
when a is more than or equal to 25.3 ℃ and less than or equal to 31.2 ℃, m (4) is [1- (a-25.3)/(31.2-25.3)]X 0.9, m (2), m (3) are all 0, m (1) [ (a-25.3) }(31.2-25.3)]×0.9，m(Θ)＝0.1；
When a is 31.2 ℃. ltoreq.a.ltoreq.35.3 ℃, m (3), m (4) are each 0, m (2) [ (a-31.2)/(35.3-31.2) ] × 0.9, m (1) [ (1- (a-31.3)/(35.3-31.3) ] × 0.9, m (Θ) [ (0.1):
when a is not less than 35.3 ℃ and not more than 39.3 ℃, m (1) and m (4) are all 0, m (3) is [ (a-35.3)/(39.3-35.3) ] × 0.9, m (2) is [1- (a-35.3)/(39.3-35.3) ] × 0.9, and m (Θ) is 0.1;
when the temperature is 39.3 ℃ and less than or equal to a, m (3) is 0.9, m (1), m (2) and m (4) are all 0, and m (theta) is 0.1.
The data collected at different times are: table 3 was obtained by using a1 ═ 30.2 ℃, 9.4klx, 438 μ L/L }, a2 ═ 33.8 ℃, 12.4klx, 358 μ L/L }, A3 ═ 38.1 ℃, 20.8klx, 283 μ L/L }, and a4 ═ 23.0 ℃, 8.5klx, 371 μ L/L }.
TABLE 3 basic credibility Allocation before combination
The above is the basic reliability calculation for the temperature, and similarly, the basic reliability calculation for the illumination intensity and the carbon dioxide volume fraction can also be completed. Wherein the indexes a, b and c respectively represent temperature, illumination intensity and volume fraction of carbon dioxide; m (1), m (2), m (3), m (4), m (Θ) represent the opening of the roll of film, opening the roll of film and starting the fan, starting the fan and the wet curtain, without action and without indeterminate support.
D-S evidence combines temperature, light intensity and temperature, and carbon dioxide volume fraction to obtain a base confidence distribution, and then combines the two sets of attributes to obtain a base confidence distribution. Finally, by a decision based on the basic confidence allocation. Determining a decision category and selecting a threshold: epsilon1＝0.2，ε20.03. Table 4 was obtained.
TABLE 4D-S evidence combinations and decisions
Wherein L (1), L (2), L (3) and L (4) are decision results, L (1) corresponds to opening the roll film, L (2) corresponds to opening the roll film and starting the fan, L (3) corresponds to starting the fan and the wet curtain and L (4) corresponds to no action.
As can be seen from table 4, through the fusion decision algorithm of the D-S evidence theory, it is concluded that the decision result can be effectively influenced if the decision result has 2 attributes, and the decision result can be determined if the decision result has 4 samples. As shown in table 1, in sample a2, after the evidence { a, b } is combined, the difference between m (1) and m (2) is higher than 0.2, m (Θ) is less than 0.03, and m (1) is much larger than m (Θ), so the decision result is L (1), { a, c } combination, on the contrary, the difference between m (2) and m (1) is higher than 0.2, m (Θ) is less than 0.03, m (2) is much larger than m (Θ), so the decision result is L (2), because the results of two-by-two combination decisions have different results, the combination needs to be further fused, and the basic confidence distribution values for m (1) and m (2) are 0.16793 and 0.81674 respectively, m (2) is higher than 0.2, m (Θ) is less than 0.03, and m (2) is much larger than m (Θ), so the decision result should be L (2); furthermore, from the variation of m (Θ) in table 3, it can be concluded that m (Θ) is greatly reduced after the fusion decision of D-S evidence theory, and after the { a, b, c } combination, the attribute reduction uncertainty of each group is reduced for the combination of 3 groups of decision attribute sets compared to the combination of 2 groups of decision attribute sets. Table 4 shows that the magnitude of uncertainty is changed from 10-2 to 10-3, so that the conclusion can be drawn that the uncertainty of a decision result can be effectively reduced by combining D-S evidences of various attribute sets, and the purpose of improving the judgment precision is further achieved.
Example 3
And selecting an SVM algorithm aiming at small sample machine learning to be compared with a decision method based on a rough set and a D-S evidence theory.
Firstly, training and testing an SVM algorithm for small sample machine learning by using the original data in the table 1, training a classifier after training and testing the original data by using the SVM algorithm, and testing a model. And evaluating the performance index of the classifier, and selecting the optimal model parameter after cross validation. The decision accuracy is 90.34%, thus the SVM algorithm of small sample machine learning can not completely learn and apply expert knowledge. According to the decision-making method based on the rough set and the D-S evidence theory, according to the reduction result of the information entropy attribute reduction, after 3 condition attributes of humidity, soil internal temperature and soil internal humidity in the table 1 are deleted, the training and the testing are carried out again, at the moment, the correct decision-making rate is increased to 100%, and the attribute reduction algorithm based on the information entropy can delete unnecessary condition attributes in expert knowledge and enhance the mapping relation between the condition attributes and the decision-making attributes. Finally, the samples A1, A2, A3 and A4 are respectively subjected to decision making by adopting a decision making method based on the rough set and the D-S evidence theory and an SVM algorithm learned by a small sample machine to obtain decision making results, the average running time of a fusion decision making algorithm based on the D-S evidence theory and the average running time of the SVM algorithm learned by the small sample machine are 0.002378S and 0.017939S respectively, and the fact that the running cost of the decision making algorithm based on the rough set and the D-S evidence theory is far less than that of the SVM algorithm learned by the small sample machine during fusion decision making is shown, so that the decision making algorithm based on the rough set and the D-S evidence theory has lower calculation complexity, and the two algorithms can process the same decision making results.
And (3) processing 12 groups of greenhouse influence factor data, comparing the processed data with expert knowledge, and judging that the identification is correct if the decision result obtained by the algorithm is consistent with the expert knowledge, and finally obtaining a correct decision result and a conclusion: the decision algorithm based on the rough set and the D-S evidence theory can distribute conflict information with less professional knowledge, and can make effective decision under the condition to obtain a correct decision result. It is feasible to apply this calculation to the control part of the greenhouse influencing factors.
In addition, the decision result of the D-S evidence theory adapts to the uncertainty of greenhouse decision factors, and the uncertainty of the decision result can be effectively reduced by adopting the D-S evidence combination of various attribute sets, so that the purpose of improving the judgment precision is achieved.
The table 1 is tested and trained by using the SVM algorithm of small sample machine learning, the obtained decision accuracy is low and is not enough to finish the condition of deciding greenhouse influence factors, the decision accuracy reaches 100% after fuzzy C-means clustering and attribute reduction algorithm based on information entropy and D-S evidence fusion are carried out again, and the efficiency and accuracy of the intelligent greenhouse decision method based on rough set and D-S evidence theory are obviously better than those of the SVM algorithm under the condition of small samples. Tables 5 and 6 are the test set data and training set data, respectively, of the SVM algorithm, as shown in the following tables:
test set data in the SVM Algorithm of Table 5
TABLE 6 training set data in SVM Algorithm
The rough set and D-S evidence theory based decision method and the short-term comparison of SVM algorithm learned by the small sample machine are shown in Table 7.
TABLE 7 greenhouse effect factor decision comparison
The above disclosure is only for a few specific embodiments of the present invention, however, the present invention is not limited to the above embodiments, and any variations that can be made by those skilled in the art are intended to fall within the scope of the present invention.
Claims (5)
1. A greenhouse intelligent decision-making method based on a rough set theory and a D-S evidence theory is characterized by comprising the following steps:
step 1: collecting greenhouse influence factor sample data, establishing a rough set and a D-S evidence theoretical model, and carrying out attribute reduction on the greenhouse influence factor sample data by taking the rough set theory as the front edge of the model;
step 2: obtaining basic confidence coefficient distribution of a combination result by using a D-S evidence combination rule and combining greenhouse influence factor sample data;
and step 3: discrete processing is carried out on the sample data by using a fuzzy C-means clustering method, and n vectors x are processedi(i-1, 2, …, n) into c fuzzy groups and finding the cluster center for each fuzzy group, minimizing the value function of the dissimilarity index, using [0,1 ] for each given data point]The degree of membership between each group is determined, and according to the normalization, the sum of the degrees of membership of one dataset is always equal to 1:
the value function of the fuzzy C-means clustering algorithm is as follows:
wherein u isijIs at [0,1 ]]C isiFor the cluster center of I in the fuzzy C-means cluster, dij＝||ci-xjI is the Euclidean distance between the I-th clustering center and the j-th data point, and m>1 is a weighted index;
and cluster center ciIs calculated as:
the membership degree is calculated as:
and 4, step 4: when the corresponding clustering category of each sample is determined, determining to use a threshold value with overlapped membership degree, and selecting all sample data corresponding to aggregation for limitation by using the threshold value of the sample data value;
and 5: constructing a basic probability distribution function by using the rough set, and calculating the support degree among all influence factors of the greenhouse;
step 6: and (3) introducing the BPA basic probability assignment matrix obtained by calculation by using an improved D-S evidence theory, and constructing a confidence matrix to complete the combination of greenhouse influence factors to obtain a decision result.
2. The intelligent greenhouse decision method based on rough set theory and D-S evidence theory as claimed in claim 1, wherein the sample data of greenhouse influencing factors collected in the step S1 comprises temperature, humidity, illumination intensity, temperature in soil, humidity in soil and volume fraction of carbon dioxide.
3. The intelligent greenhouse decision method based on rough set theory and D-S evidence theory as claimed in claim 1, wherein the decision result mentioned in the step S6 includes four categories of 1, 2, 3 and 4, which respectively correspond to the operations of opening the rolling film, opening the rolling film and starting the fan, starting the fan and the wet curtain and not acting.
4. The intelligent greenhouse decision making method based on rough set theory and D-S evidence theory as claimed in claim 3, wherein the decision result is set as { L (k) }, where k is the result of 4 decisions, and k is 1, 2, 3, 4, and the greenhouse influence factor control decision identification boxThe basic credibility distribution function under the power set of the frame represents the support degree of greenhouse influence factor information of various decision categories, and
∑A∈Θm (a) ═ 1, where m (1), m (2), m (3), m (4) represent the basic credibility distribution of greenhouse decision factors, and m (Φ) represents the basic credibility allocation of uncertain greenhouse influencing factors.
5. The intelligent greenhouse decision method based on rough set theory and D-S evidence theory as claimed in claim 4, wherein the basic confidence distributions M (1), M (2), M (3), and M (4) of the greenhouse decision factors correspond to a set of temperature values {31.5 ℃, 32.8 ℃, 29.2 ℃ }, {35.1 ℃, 36.0 ℃, 34.7 ℃ }, {38.2 ℃, 42.3 ℃, 37.5 ℃ }, {24.9 ℃, 27.0 ℃, 25.1 }.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110150119.7A CN112785004A (en) | 2021-02-03 | 2021-02-03 | Greenhouse intelligent decision-making method based on rough set theory and D-S evidence theory |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110150119.7A CN112785004A (en) | 2021-02-03 | 2021-02-03 | Greenhouse intelligent decision-making method based on rough set theory and D-S evidence theory |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112785004A true CN112785004A (en) | 2021-05-11 |
Family
ID=75760731
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110150119.7A Pending CN112785004A (en) | 2021-02-03 | 2021-02-03 | Greenhouse intelligent decision-making method based on rough set theory and D-S evidence theory |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112785004A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113689114A (en) * | 2021-08-23 | 2021-11-23 | 中国工商银行股份有限公司 | Method, device and equipment for determining credit degree |
| CN116737940A (en) * | 2023-08-14 | 2023-09-12 | 成都飞航智云科技有限公司 | Intelligent decision method and decision system |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102937785A (en) * | 2012-11-09 | 2013-02-20 | 江苏大学 | Greenhouse multi-model information fusion based methods and system for regulating and controlling greenhouse environment |
| CN107644267A (en) * | 2017-09-11 | 2018-01-30 | 河南科技大学 | A kind of green house control Decision fusion method based on D S evidence theories |
| CN111625030A (en) * | 2020-05-19 | 2020-09-04 | 北京工业职业技术学院 | Greenhouse environment control method, device, equipment, system and storage medium |
-
2021
- 2021-02-03 CN CN202110150119.7A patent/CN112785004A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102937785A (en) * | 2012-11-09 | 2013-02-20 | 江苏大学 | Greenhouse multi-model information fusion based methods and system for regulating and controlling greenhouse environment |
| CN107644267A (en) * | 2017-09-11 | 2018-01-30 | 河南科技大学 | A kind of green house control Decision fusion method based on D S evidence theories |
| CN111625030A (en) * | 2020-05-19 | 2020-09-04 | 北京工业职业技术学院 | Greenhouse environment control method, device, equipment, system and storage medium |
Non-Patent Citations (2)
| Title |
|---|
| 刘曙光 等: "《模糊控制技术》", 中国纺织出版社, pages: 31 - 32 * |
| 王俊 等: "融合粗糙集和证据理论的温室环境控制推理决策方法", 《农业工程学报》, vol. 28, no. 17, pages 172 - 178 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113689114A (en) * | 2021-08-23 | 2021-11-23 | 中国工商银行股份有限公司 | Method, device and equipment for determining credit degree |
| CN116737940A (en) * | 2023-08-14 | 2023-09-12 | 成都飞航智云科技有限公司 | Intelligent decision method and decision system |
| CN116737940B (en) * | 2023-08-14 | 2023-11-07 | 成都飞航智云科技有限公司 | Intelligent decision method and decision system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Aïvodji et al. | Fairwashing: the risk of rationalization | |
| CN110213222B (en) | Network intrusion detection method based on machine learning | |
| CN111324642A (en) | Model algorithm type selection and evaluation method for power grid big data analysis | |
| CN107292097B (en) | Chinese medicine principal symptom selection method based on feature group | |
| CN111612261B (en) | Financial big data analysis system based on block chain | |
| CN112001788B (en) | Credit card illegal fraud identification method based on RF-DBSCAN algorithm | |
| CN112785004A (en) | Greenhouse intelligent decision-making method based on rough set theory and D-S evidence theory | |
| CN111046930A (en) | Power supply service satisfaction influence factor identification method based on decision tree algorithm | |
| CN111539451A (en) | Sample data optimization method, device, equipment and storage medium | |
| Yang et al. | An Evidence Combination Method based on DBSCAN Clustering. | |
| CN113516189B (en) | Website malicious user prediction method based on two-stage random forest algorithm | |
| Zhang et al. | Dbiecm-an evolving clustering method for streaming data clustering | |
| CN109685133A (en) | The data classification method of prediction model low cost, high discrimination based on building | |
| CN113591947A (en) | Power data clustering method and device based on power consumption behaviors and storage medium | |
| Halkidi | Quality Assessment and Uncertainty Handling in Data Mining Process. | |
| CN111683141A (en) | User demand-oriented dynamic QoS service selection method and system | |
| CN116485021A (en) | Coal enterprise technical skill person sentry matching prediction method and system | |
| CN115908256A (en) | Yarn quality prediction method based on meta-learning method | |
| CN115600913A (en) | Main data identification method for intelligent mine | |
| CN112288571A (en) | Personal credit risk assessment method based on rapid construction of neighborhood coverage | |
| Duan et al. | Risk assessment for enterprise merger and acquisition via multiple classifier fusion | |
| Bouhamed et al. | Data and information quality assessment in a possibilistic framework based on the Choquet Integral | |
| Zhao et al. | Outlier detection for partially labeled categorical data based on conditional information entropy | |
| Ma et al. | An equidistance index intuitionistic fuzzy c-means clustering algorithm based on local density and membership degree boundary | |
| CN116701962B (en) | Edge data processing method, device, computing equipment and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |