Next Article in Journal
Language Models for Everyone—Responsible and Transparent Development of Open Large Language Models
Previous Article in Journal
Preface: The 3rd International Day on Computer Science and Applied Mathematics (ICSAM’23)
 
 
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Proceeding Paper

Machine Learning Explicit and Implicit Model for Factor Classification Based on Factor Space Theory †

1
College of Science, Liaoning Technical University, Fuxin 123000, China
2
Institute of Intelligent Engineering and Mathematics, Liaoning Technical University, Fuxin 123000, China
*
Author to whom correspondence should be addressed.
Presented at the 2023 Summit of the International Society for the Study of Information (IS4SI 2023), Beijing, China, 14–16 August 2023.
Comput. Sci. Math. Forum 2023, 8(1), 88; https://doi.org/10.3390/cmsf2023008088
Published: 4 September 2023

Abstract

:
Factor implicit thought in factor space theory can be used to solve the classification problem of machine learning. On the basis of factor implicitness, a serial sweeping class classification algorithm is proposed, and a factor implicit model is constructed with this algorithm, and then tested and classified. On the basis of a serial scanning algorithm, this paper proposes a fine-tuning Sweeping Learning Algorithm, an ascending Side-By-Side Serial Scanning Algorithm, and a combination algorithm to solve the two problems in the running of the serial scanning algorithm. In order to solve the problem that the training speed of a traditional serial scanning algorithm is too slow due to the large amount of data, a new method which can select mixed domains in advance—the partial-side serial scanning algorithm.

1. Introduction

As the core of artificial intelligence, machine learning plays a very important role in the realization of computer intelligence. With the continuous innovation of science and technology and the rapid development of computer network technology, machine learning has become an increasingly important part of the field of artificial intelligence, and the classification problem in machine learning is also one of the main tasks of machine learning, which is widely used in various fields of real life, and the accuracy rate of the goals it can achieve is getting higher and higher. Among them, the binary classification problem is also an important part of machine learning; whether it is medical, agriculture or daily production and life, the problem of binary classification is everywhere. Looking for a more accurate algorithm to solve the binary classification problem is an important research direction in the field of artificial intelligence.
The advent of the information revolution and the era of big data promoted the development of artificial intelligence, followed by the need to find how to save, extract, and process the required factor data for huge data. Causal analysis between factors provides an important tool for artificial intelligence, data mining, etc. However, the main difficulty in the field of artificial intelligence is that the key factors to solve practical problems have not been revealed, and how to find the key factors has become an important research direction. Factor explicitness is a new theory under the factor space theory proposed by Wang Peizhuang [1]. As a bottleneck problem in the field of artificial intelligence, factor explicitness has great significance in helping artificial intelligence problems find the key factors. As long as the key factors are found, the corresponding problems will naturally be solved. Sun Hui et al. [2] proposed a serial sweep algorithm. Aiming at the classification problem of machine learning, the algorithm defines the sweeping direction and the explicit and implicit factors by using the factor space theory. In order to reduce algorithm’s complexity, the ordered set of swept class vectors is defined, and the factor implicit model is constructed. The results of numerical experiments show that the algorithm is feasible and effective. Zeng Fanhui et al. [3] proposed the application of the serial scanning algorithm in multi-classification. On the basis of this algorithm, this paper proposes a fine-tuning sweeping learning algorithm, a dimension-raising side-by-side serial scanning algorithm, and a combination algorithm [4] to solve the two problems in the operation of the serial sweep classification algorithm.

2. Sweeping Learning Algorithm

(1)
Algorithm steps:
Algorithm steps:
1:
input S: = {x1, …, xI}; S+: = {x1+, …, xJ+};
2:
w: = o+o;
3:
l: = max{(xI, w)}; u: = min{(xi+, w)}; r: = (ul)/2; o: = (u + l)/2;
4:
if l < u, then go to Step 3; otherwise:
    S: = S − {xIS|(xI, w) < u};
    S+: = S+ − {xj+S+|(xj+, w) > l};
    go back 2.
5:
for each deleted xI from S, if l < (xI, w) < u, then S: = S + {xI};
for each deleted xj+ from S+, if l < (xj+, w) < u, then S+: = S+ + {xj+}.
    go back 2.
6:
if x is all deleted from S + S+, (x, w) never enter into (l, u); then, output w, which can divide two classes’ points.
(2)
Fine-tuning algorithm step
Fine-tuning algorithm step
1:
calculate the sum of two types of centers u t and u t + and obtain wt;
2:
take any integer t = [0, T); it is best to select the serial classification algorithm before sweeping into infinite iteration; then, adjust the displacement of the remaining positive and negative mixed points class X t + 1 + , X t + 1 to fine-tune the positive and negative class data set.
3:
use linear discriminant analysis (LDA) to solve w*, wt+1 = w*.
4:
starting from t = t + 1, calculate solve, stop, and output the explicit and implicit factors by using the scanning serial classification algorithm.
(3)
Steps of dimension-raising algorithm
Steps of dimension-raising algorithm
1:
calculate the sum of two types of centers u t and u t + and obtain wt;
2:
Take any integer t = [0, T); it is best to select the serial classification algorithm before sweeping into infinite iteration; then, adjust the displacement of the remaining positive and negative mixed points class X t + 1 + , X t + 1 to fine-tune the positive and negative class data set.
3:
use the kernel function to raise the dimension of the fine-tuned positive and negative data set, and use the corresponding formula to perform projection calculation.
4:
Starting from t = t + 1, calculate, solve, stop, and output the explicit and implicit factors using the sequential classification algorithm of sweeping classes.
(4)
Example application
The data of Sweeping Learning Algorithm are derived from the Cryotherapy data set, Immunotherapy data set and Somerville Happiness in the UCI database based on the survey data set; Matlab software is used to show that the serial scanning algorithm has a better classification effect compared with a support vector machine, and the experimental results also show the feasibility and practicability of the serial scanning algorithm. The experimental results, such as related data set physical properties and experimental results, are shown in Table 1, Table 2, Table 3 and Table 4.

3. Side-By-Side Serial Scanning Algorithm

(1)
Algorithm steps:
Algorithm steps:
1:
input S: = {x1, …, xI}; S+: = {x1+, …, xJ+};
2:
w: = o+o; l: = max {(xI, w)}; u: = min {(xi+, w)}; r: = (ul)/2; o: = (u + l)/2;
3:
if l < o < u, then goto 4; Else
S: = S − {xIS|(xI, w) < u};
S+: = S+ − {xj+S+|(xj+, w) > l};
go back 2.
4:
for each deleted xI from S, if l < (xI, w) < u, then S: = S + {xI};
for each deleted xj+ from S+, if l < (xj+, w) < u, then S+: = S+ + {xj+};
go back 2.
5:
if x is all deleted from S + S+, (x, w) never enter into (l, u); then, output w, which can divide two classes’ points.
Attachment: In the process of program calculation, the size of the sample centers of the positive and negative class data will change in actual situations, so it is difficult to maintain the positive class center above the negative class center; thus, it is necessary to consider the position transformation of the two types of centers. When it appears below, the calculation method of class vector scanning is: w t = o t + + o t .
(2)
Example application:
Using the data sets of Iris and Haberman’s Survival UCI, SVM, sweep, s-sweep, and matlab2018a coding were used for multiple experimental comparisons. In the experimental process, 80% of the data sets were randomly selected for testing and the remaining 20% were verified. The corresponding attributes of the six data sets shown in Table 5 and Table 6 are the average values of multiple experimental results.

4. Summary and Prospects

This paper solves two problems in the operation of a serial class scanning algorithm and puts forward corresponding processing methods. By contacting the vertical bisector of the class scanning vector, the sample points are retained according to the distance between the projection distance of the sample points in the class scanning vector and the vertical bisector, and a new training sample is obtained. New training samples are used to replace original training samples for traditional serial scanning training, as shown in Table 6. This algorithm can reduce the training samples without affecting the classification ability of the class scanning vector and speed up the data-processing process. Meanwhile, when the positive class center is smaller than the negative class center in the process of class scanning, the influence of the obtained class scanning vector on the data prediction and classification is also a problem. At the same time, the simulation results show that the algorithm is effective and feasible, and it has faster data-processing ability and accuracy than a traditional SVM algorithm and serial scanning algorithm.

Author Contributions

There are five authors in this paper. K.Z. provided the algorithm and software coding for verification analysis and writing of the paper; F.Z. provided the guidance for writing and preparing the first draft; X.L., K.L. and Y.W. reviewed and edited the paper. All authors have read and agreed to the published version of the manuscript.

Funding

This research was conducted by Fuxin Social Science Project: Mathematical Basic Research of Factor Space Theory under the background of Fuxin Business Environment Optimization, project number: 2022Fsllx111; Basic Scientific Research Project of colleges and universities of Liaoning Provincial Department of Education, key research project: Theory and Application Research of Factor Space-based intelligent incubation under the Digital Background, project number: funded by LJKZZ20220047.

Institutional Review Board Statement

Not applicable.

Informed Consent Statement

Not applicable.

Data Availability Statement

The experimental data in this paper are derived from the data set of UCI machine learning repository.

Conflicts of Interest

The authors declare no conflict of interest.

References

  1. Wang, P.Z.; Zeng, F.H. Factor Space Theory, Mathematical Basis of Unified Intelligence Theory; Science Press: Beijing, China, 2023. [Google Scholar]
  2. Sun, H. Sweeping Chain Learning Algorithm. Master’s Thesis, Liaoning Technical University, Fuxin, China, 2022. [Google Scholar]
  3. Zeng, F.H.; Wang, Y.; Wang, P.Z.; Sun, H. Multi-classes sweep learning based on factor space theory. J. Liaoning Tech. Univ. Nat. Sci. 2022; accepted. [Google Scholar]
  4. Zhang, K.; Zeng, F. Sweeping class linkage algorithm in factor space. J. Intell. Syst. 2023; submitted. [Google Scholar]
Table 1. Attributes for the two data sets.
Table 1. Attributes for the two data sets.
Data SetSample NumberNumber of CategoriesCharacteristic Number
Cryotherapy9026
Immunotherapy9027
The Somerville Happiness Survey14326
Table 2. Performance comparison of Cryotherapy data sample set.
Table 2. Performance comparison of Cryotherapy data sample set.
Contrast TermNumber of Training StepsNumber of SamplesAccuracy %
SVM749100%
sweep749100%
Table 3. Performance comparison of Immunotherapy data sample set.
Table 3. Performance comparison of Immunotherapy data sample set.
Contrast TermNumber of Training StepsNumber of SamplesAccuracy %
SVM801090%
sweep801090%
Table 4. Performance comparison of the Somerville Happiness Survey data sample set.
Table 4. Performance comparison of the Somerville Happiness Survey data sample set.
Contrast TermNumber of Training StepsNumber of SamplesAccuracy %
SVM1331070%
Sweep13310100%
Table 5. Attributes for the two data sets.
Table 5. Attributes for the two data sets.
Data SetSample NumberNumber of CategoriesCharacteristic Number
Iris15034
Haberman’s Survival30623
Table 6. Performance comparison on Iris data sample set.
Table 6. Performance comparison on Iris data sample set.
Contrast TermNumber of Training StepsNumber of SamplesTraining Time/msAccuracy %
SVM10050259.15696.5%
sweep100507.369100%
s-sweep100501.987100%
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Zhang, K.; Zeng, F.; Liu, X.; Lin, K.; Wang, Y. Machine Learning Explicit and Implicit Model for Factor Classification Based on Factor Space Theory. Comput. Sci. Math. Forum 2023, 8, 88. https://doi.org/10.3390/cmsf2023008088

AMA Style

Zhang K, Zeng F, Liu X, Lin K, Wang Y. Machine Learning Explicit and Implicit Model for Factor Classification Based on Factor Space Theory. Computer Sciences & Mathematics Forum. 2023; 8(1):88. https://doi.org/10.3390/cmsf2023008088

Chicago/Turabian Style

Zhang, Kaijie, Fanhui Zeng, Xiaotong Liu, Kaile Lin, and Ying Wang. 2023. "Machine Learning Explicit and Implicit Model for Factor Classification Based on Factor Space Theory" Computer Sciences & Mathematics Forum 8, no. 1: 88. https://doi.org/10.3390/cmsf2023008088

Article Metrics

Back to TopTop