Fuzzy sets for data mining and recommendation algorithms

2.1 Related work

Since many factors influence the stock price, it has a non-linear structure, so trading in forecasting and recommending exchanges is a very difficult task [15, 16]. In a user-friendly way, Wang proposed a new forecasting system and stock trading advice. The recommendation system can inform the user about the stock trading information in the next step, and use the detailed information to convert the initial time series into meaningful and the explained particles also provide a more efficient way to predict heterogeneous parts. The system first determines the interval based on the detailed information, determines the fuzzy set and erodes the historical data, then establishes the fuzzy relationship and assigns weights for each time period, and finally performs prediction and recommendation [17]. The experimental results show that the system has the best predictive index and increases the possibility of profit. Zhao Xiaoqiang applied vector fuzzy support machine (FSVM) data mining to training sets with large numbers of samples and classification slow classification defects. Vector algorithm is used in the improved data mining fuzzy classifier. It is the first efficient algorithm to support candidate vectors, which reduces the size of the training set to improve the learning speed. Secondly, a new fuzzy membership function is proposed. The hyperplane effect of the auxiliary vector on the optimal classification of fuzzy auxiliary vector is used, and the pre-selected training set is used for learning [18, 19]. Finally, the particle swarm optimization algorithm is used to select the best subset of the support vector, and the average classification error is obtained. Used as a fitness function. When the last particle is output, the membership in the sample is compared to the set threshold, and the support vector with a relatively high membership is selected as the new reference vector to improve the classification speed. The experimental results show that the proposed algorithm improves the learning speed and classification speed of fuzzy support vectors without loss of classification accuracy.

2.2 Fuzzy set

(1) Principle of fuzzy set

In fact, collections are highly abstractions of certain objective things, and fuzzy sets S must be a collection of such objects: for any object, in addition to two possible possibilities, it still exists. “Can determine exactly whether object O belongs to set S” or “to some extent, object O belongs to set S”, this “belonging” state is between “fully owned” and “completely not owned”, this It is the ambiguity of the set S. Since the inherent characteristics of human knowledge are its incompleteness, uncertainty and inconsistency, this ambiguity can be seen anywhere in the data source. On the one hand, in general data sources, the number of values of the quantity attribute is very large, it is difficult to accurately find the relationship between different value points or value ranges of each attribute, which is not necessary, so people usually pay attention to the other side. Due to the diversity of the objective world and complexity, people’s understanding of many things is based on fuzzy concepts, so we must use fuzzy concepts to abstract and generalize some attributes of events. Zade’s set theory can introduce the concept of ambiguity for specified attributes when retrieving data sources. Used to convert attribute values, so the attribute values in the data source are easy for human understanding and computer analysis and treatment. For ease of description, this article is for each attribute. It should be noted that when determining the membership function, it should be determined according to the actual situation of the problem. The fuzzy subset A in this universe means that for any x∈A, a quantity μA(x) is defined, where μA(x) is x Members in A, while μA(x)∈[0,1]; mapping X⟶[0,1], x⟶μA(x) is called membership function A, and fuzzy subsets are often called fuzzy sets or Fuzzy set. In the moving target detection application, each attribute is a fuzzy set, and any fuzzy set corresponds to a membership function. Therefore, in order to truly describe various attribute values, it is very important to select an appropriate membership function model.

2.3 Research on data mining and recommendation algorithm

(1) Introduction of data mining

When studying knowledge engineering, there are always problems such as information ambiguity. There are three kinds of ambiguity: term ambiguity (such as high and low), data uncertainty (such as noise), and knowledge uncertainty (such as rules before and after rules) between the dependencies). One of the basic theories of artificial intelligence is not enough to solve these uncertainties. In this paper, fuzzy set theory is used to transform fuzzy attributes, and fuzzy association rules in data sources are obtained. A fuzzy association rule retrieval algorithm is proposed and applied to refining. One aspect of the industry can further explore the role of fuzzy set theory in data mining to improve the quality of data mining. In recent years, the application of fuzzy set theory in data mining research has become a hotspot in data mining research. In order to track the progress of research and study the future research direction, we consider the main research directions of fuzzy set theory in data mining (cluster analysis, association analysis, classification), mainly to explain the representation of data and patterns, and the calculation mode. Key issues such as similarity can be seen in the use of powerful fuzzy data modeling functions in fuzzy data theory, combined with other intelligent processing techniques is the main technology in this field. Data mining solves the problem of further developing knowledge-based intelligent systems because of the lack of knowledge and the difficulty of acquiring knowledge.

(2) The development of data mining

As the total amount of information continues to grow, there is an urgent need for an effective information analysis tool that can detect hidden relationships between large amounts of data and extract useful information knowledge from large amounts of data. Despite the emergence of simple data statistics methods, modern intelligent data analysis tools are not yet mature. Therefore, there is a big gap between data generation and understanding. Data mining is a new type of data mining technology that can solve this controversy. The purpose of data mining is to extract hidden prediction information from large data sources (also known as discovery knowledge), which can explore potential patterns between data and identify information that business operators may miss in the form of understanding and observation. An important aspect of data mining applications is the identification of conventional, new, useful, and understandable patterns from large amounts of raw data. Over the past few decades, fuzzy set theory has been successfully used for pattern recognition, intelligent control, and machines. Learning, artificial intelligence and other areas of research. For a long time, the expression of knowledge and the rationality of knowledge have always been the main direction of research in the field of fuzzy set theory. The result provides a theoretical basis for the construction of knowledge-based intelligent design systems. In the face of a large amount of data, it cannot be collected manually. When sorting and using traditional data analysis processing tools to acquire knowledge to solve this problem, the emergence of data mining has opened up a new way for knowledge acquisition. It adapts to other technologies such as databases, intelligent learning, and artificial development technologies, and draws useful knowledge from it, greatly severing the problem of knowledge-based intelligent systems.

(3) Recommendation algorithm

With the creation and development of fuzzy sets, fuzzy reasoning has become an important part of the field of computational intelligence. Regarding the main problems of fuzzy thinking, scientists at home and abroad have conducted extensive research and proposed various fuzzy inference algorithms, which are all intelligent algorithms. Reasoning is one of the main characteristics of human intelligence, and an important technology to realize artificial intelligence. With the creation and development of fuzzy sets, fuzzy reasoning has become an important part of the field of computational intelligence. Fuzzy thinking aims to model human thinking models and build incorrect thinking methods. The reasoning process is based on rules and given facts, so people can ask questions and a rough answer based on incorrect and imperfect knowledge base. Currently, fuzzy inference technology is widely used in many fields of intelligent systems, such as fuzzy control systems, fuzzy expert systems, fuzzy neural network systems and fuzzy decision support systems.

2.4 Research on data mining and recommendation algorithm of fuzzy sets

(1) In the excavation process of any scheme I, if the support speed is greater than or equal to the minimum support speed, the support speed of the non-empty sub-mode must also be greater than or equal to the minimum support speed, which is to check the candidate template, reduce the number of candidate templates, and speed up the excavation speed. And plays an important role in reducing storage capacity.

(2) Algorithm Description: First, initialize the data structure of the frequent mode, use the Match function to find all frequent patterns of length k, use the Candidates function to generate the appropriate frequent patterns, and associate all fuzzy attributes of the pattern with the rules. Moreover, the algorithm can be improved and simplified, that is, when the frequent template is generated for the first time, it is only used for searching and generating in the first attribute, and when the candidate frequent pattern is generated for the kth time, only the kth frequent candidate mode and the first k, combining+1 attributes to create possible candidate templates, and not using any other attributes after the k + 1 attribute, which can significantly speed up the mining and reduce the number of possible templates.

(3) Algorithm example

Definition one: Fuzzy mode P = A 1 ∩ A 2 ∩ · · · ∩ A k(1)

The length of the mode P is called k, and AK is called the item: A i = ( a i : t i ) a i ∈ A , t i ∈ v i ( i = 1 , 2 , · · · , m )(2)

The mode consisting of any q(q < k) items from A 1 , A 2 , . . . , A k is called the q sub-mode of P. Definition two: Extend the project Ai, A i = ( a i ⩾ t i )(3)

Indicates that the membership of the fuzzy attribute ai is greater than or equal to ti. This definition plays an important role in mining higher-level association rules.

Definition three:

1) Take the fuzzy sample of the data source, calculate the mutual information of each feature according to the fuzzy value, and get the exact value.

2) Select the maximum mutual information feature Ak;

3) The example of taking the same value at Ak is divided into the same sample subset, and the category result of the sample subset is fuzzy.

4) A recursive call-building algorithm for a subset of multiple fuzzy samples.

5) Limit the fuzzy category.

If the subset contains only the example of the fuzzy category, the corresponding decision tree is marked with C1, C2, C3, and returns to the call. The calculation process of the fuzzy algorithm is as follows:

1) Calculating information entropy entropy H (U) H ( U ) = - ∑ i P ( u i ) log 2 p ( u i )(4)

Among them, P ( u i ) = | u i | / | S | . Indicates the total number of training sample subsets S, |u_i|indicates the number of instances where the sample ambiguity value is a category.

2) Conditional entropy H (U ∖ V)

H ( U ∖ V ) = - ∑ i P ( v i ) ∑ i P ( u i | v i ) log 2 P ( u i | v i )(5)

When the attribute Ai takes the value v_i, the conditional probability of the category u_i is: P ( u i | v i ) = | u i | / | v i |(6)

3) Calculate mutual information. I ( A 1 ) = H ( U ) - H ( U | V )(7)

4) Choose larger I(A_i), Recursively establish a decision tree.

Definition IV: In order to solve the shortcomings of the CRI method, Wang Guojun proposed the full implication of the fuzzy inference from the perspective of logical semantic implication. The core of this algorithm is the three I principle of fuzzy reasoning.

The solution B* of the FMP problem is the smallest fuzzy set in the domain Y that makes the following formula take the maximum value: ( A ( x ) → B ( y ) ) → ( A * ( x ) → B * ( y ) )(8)

The solution to the FMT problem A * is the largest fuzzy set in the domain X that makes the above formula take the maximum value.

The solution B^* of the FMP problem is the largest fuzzy set in the domain Y that makes the following formula take the maximum value. ( A * ( x ) → B * ( y ) ) → ( A ( x ) → B ( y ) )(9)

3) y The ∂-B* of the FMP problem is the largest fuzzy set in the domain y that makes the following formula hold: ( A * ( x ) → B * ( y ) ) → ( A ( x ) → B ( y ) ) ⩾ ∂(10)

4) Turksen give a formula for calculating similarity: S AARS ( A , A * ) = ( 1 + DM ( A , A * ) ) - 1(11)

3.1 Experimental setup

(1) Data collection

The concept of fuzziness comes from the fuzzy linearity unique to the display world. It makes full use of the powerful fuzzy data modeling function of fuzzy data theory and combines it with other intelligent processing methods. Many algorithms for intelligent analysis of fuzzy association rules are proposed. Fuzzy computing is based on fuzzy set theory, which can simulate the inexact and non-linear ability of human brain processing information and is used in many fields. On the other hand, fuzzy dictionary is used to describe the numerical values and general properties of relational databases, and then a kind of the algorithm is used to develop fuzzy association rules for mixed data, which effectively eliminates the obstacles caused by knowledge bottlenecks and obtains the knowledge connotation. The intelligent system further promotes the development of technology, makes the quality more efficient, and makes the decision tree of mining results more scientific.

(2) Experimental steps

Determination of fuzzy sets.

3.2 Experimental process

(1) Using data collection algorithms and fuzzy set recommendations as the main techniques, using dedicated input and output devices to measure the economics of the plant, and then comparing the data. For example, an oil refinery prepares an appropriate production plan based on the physical properties of the crude oil and the conditions of the refinery, and prepares and adjusts the production and refining plan based on the actual conditions of the existing crude oil, and then estimates which type of crude oil to use, and which refining plan. It can provide high product productivity and high economic efficiency, calculate production efficiency before and after fuzzy data mining algorithms and recommendations, and the yield will change slightly. Through the comparison and improvement of experiments, the data comparison model is used before and after. The results of fuzzy data mining are more vivid and vivid, which provides powerful technical support for data mining efficiency, improves the quality of data mining, and makes the results more professional.

4.1 Comparison of data analysis

(1) Fis refers to a fuzzy inference system, also known as a fuzzy system, which has the ability to process fuzzy information based on methods such as fuzzy set theory and fuzzy inference. The structure of the fuzzy set depends on two aspects: the corresponding domain and the corresponding membership function. The description of the membership function is subjective, which means that different people describe the same concept, and the resulting membership function may be very different. As shown in Table 2 of Fig. 1.

OPEN IN VIEWER

Fig.1

Production speed and output table.

(2) Data mining is the process of finding target data from massive data. Massive data has a large number of interference options, with fuzzy features and randomness, the location of the target data is unknown, hidden in massive data and the category is unknown, so it is difficult to completely find the target data using traditional data filtering methods. Data analysis and fuzzy set based algorithms facilitate them. Among them, D1 and D2 are data mining, and the fuzzy set is gradually used in the middle. Before using D3, you can see that the improved algorithm in KDD database is very short, can handle small CUP2011 sub-database, and the detection accuracy is higher, reaching 95% feasible. Sex, showing an improved test result algorithm, as shown in Table 3 of Fig. 2.

OPEN IN VIEWER

Fig.2

Fuzzy set algorithm test result.

(1) Data mining refers to extracting unknown, valuable models or laws from large amounts of data as the total amount of information increases. Despite the simple method of data statistics, modern intelligent data analysis tools are not yet mature. Due to the urgent need for effective information analysis tools, a large number of hidden relationships between data are found, and useful information or knowledge is extracted from a large amount of data. Then there is the fuzzy set mining. Extract legal, new, useful, and easy-to-understand templates from large amounts of raw data. At the same time, it is also widely used in the refining industry. After gradually introducing data mining algorithms and fuzzy set recommendations, the factory sales changes are shown in Fig. 3.

OPEN IN VIEWER

Fig.3

Sales record.

(2) In fact, large data sets have complex data relationships, highly nonlinear and ubiquitous noise data, which makes the application of some algorithms very limited and cannot effectively express the relationship between data. This has certain limitations. In order to study and implement the fuzzy data mining algorithm, the fuzzy dial is used to represent the relationship between data, and the data source is transformed to expand the range of representation rules. The system has higher prediction performance, improved classification speed, and reduced Storage capacity, etc. In the case of random surveys of 100 people of different ages, the fuzzy data analysis algorithms and suggestions are satisfactory. The survey results are shown in Fig. 4.

OPEN IN VIEWER

Fig.4

Fuzzy set data mining and recommendation algorithm satisfaction survey.