A fast classification method of mass data in Internet of things based on fuzzy clustering maximum tree algorithm

Abstract

In order to improve the classification accuracy and shorten the classification time of mass data, a fast classification method of mass data in the Internet of things based on fuzzy clustering maximum tree algorithm is proposed. Reduce the dimension to process the mass data of the Internet of things, establish the time series of the mass data of the Internet of things, and complete the preprocessing of the mass data of the Internet of things. Extract the feature vector of the Internet of things mass data, and use the fuzzy clustering maximum tree algorithm to perform fuzzy clustering analysis on the Internet of things mass data, so as to realize the classification of the Internet of things mass data. The results show that the recall rate of the proposed method is as high as 97.5%, the root mean square error is only 0.030, and the classification time is only 12.3 ms.

Keywords

Fuzzy clustering maximum tree algorithm principal component analysis Internet of things mass data rapid classification

1. Introduction

In recent years, the rapid development of the Internet of things and mobile Internet has injected new vitality into various industries. The era of intelligence based on big data has arrived. In this era of interconnection of all things, data inevitably begins to grow in geometric progression, the speed of data generation is faster and faster, and the amount of data increases explosively [5,11,13]. Especially in the Internet of things, in the environment of interconnected things, mass data is complex and the transmission process is dynamic and unstable. Therefore, a large number of different RFID readers and sensors must be deployed to the Internet of things environment to monitor the mass data generated in the Internet of things in real time. In the era of big data, mass IOT data contains infinite value. How to mine the value of mass IOT data has become a new industry hotspot [1,3,20]. Especially in the Internet of things, facing tens of thousands of mass data, how to effectively classify and extract mass data is a very important research topic. Therefore, it is of great significance to quickly classify the mass data of the Internet of things.

At present, scholars in related fields have carried out research on the classification of mass data of the Internet of Things. Reference [10] proposed a classification method for noisy and incomplete IoT datasets based on time series data analysis. This paper presents results using topological data analysis to understand a dataset of hundreds of interacting IoT devices located in multiple residential environments captured over 9 months. Data sets are noisy, incomplete, and subject to fluctuations in multiple life patterns. We treat the dataset as a collection of multi-attribute time series and conduct several types of IoT classification experiments. The results are compared with other single-attribute and multi-attribute techniques for time series analysis. The results show that topological data analysis does a particularly good job of classifying incomplete, noisy, and life-pattern-related IoT data compared to these other standard methods. Reference [6] proposed a temporal distribution feature learning method in IoT network traffic classification. Based on a new representation of network data, traffic data is treated as a series of images. Therefore, network data is implemented as a video stream to employ temporal distribution feature learning. Temporal information in network statistics is learned using convolutional neural networks and long short-term memory, and pseudo-temporal features between streams are learned using temporally distributed multilayer perceptrons. Experiment with a larger dataset with more classes. Experimental results show that temporal distribution feature learning improves network classification performance by 10%. Reference [8] proposes an energy- and trust-aware big data classification and secure routing algorithm based on MapReduce framework in the Internet of Things. The proposed model is based on the traditional energy harvesting trust-aware routing algorithm and is designed with a link lifetime model. The model utilizes cost metric and comprehensively considers factors such as delay, link life, energy and trust, and effectively selects the optimal safe routing path. The big data classification process is performed at the base station using the MapReduce framework. Therefore, big data classification is performed using stacked autoencoders trained by the adaptive E-Bat algorithm. The adaptive E-Bat algorithm is developed by combining the adaptive concept with the Bat algorithm and exponentially weighted moving average. The proposed model energy harvesting trust-aware routing algorithm exhibits better performance by obtaining a maximum energy of 0.9855. However, the above methods still have the problems of poor classification effect, low accuracy and long time.

To solve the above problems, this paper proposes a fast classification method for mass data in the Internet of things based on fuzzy clustering maximum tree algorithm. The specific technical route is as follows:

Step 1: use the principal component analysis method to reduce the dimension and process the mass data of the Internet of things. Based on the interval number theory, the time series of mass data in the Internet of things is established.

Step 2: select the feature items of the mass data of the Internet of things, build the feature vector space model of the mass data of the Internet of things, and form the feature vector of the mass data of the Internet of things.

Step 3: use the fuzzy clustering maximum tree algorithm to perform fuzzy clustering analysis on the mass data set of the Internet of things to realize the classification of mass data of the Internet of things.

Step 4: experimental analysis

Step 5: conclusion.

2. Design of fast classification method for mass data in Internet of things

2.1. Mass data preprocessing of Internet of things

In order to speed up the classification of the mass data of the Internet of Things and improve the classification accuracy of the mass data of the Internet of Things. Using the unsupervised principal component analysis method [4,12,16], the mass IoT data is mapped to the obtained optimal subspace of IoT mass data distribution, and the dimensionality reduction processing of the IoT mass data is completed.

Assuming that the IoT mass data set consisting of q IoT mass data features w is $W = {w_{1}, w_{2}, \dots, w_{q}}$ , the expected values of the e feature $w_{e}$ and the $e + 1$ feature $w_{e + 1}$ are $exp (w_{e})$ and $exp (w_{e + 1})$ , respectively. The formula for calculating the covariance between the two IoT mass data features is as follows: $\begin{matrix} (1) & cov (w_{e}, w_{e + 1}) = exp [\frac{w_{e} - exp (w_{e})}{w_{e + 1} - exp (w_{e + 1})}] \end{matrix}$

From this, the attribute covariance matrix [2] expression of the mass IoT dataset W is derived, as follows: $\begin{matrix} (2) & cov (W) = [\begin{matrix} cov (w_{1}, w_{1}) & \dots & cov (w_{n}, w_{1}) \\ \dots & \dots & \dots \\ cov (w_{1}, w_{e}) & \dots & cov (w_{n}, w_{n}) \end{matrix}] \end{matrix}$

Orthogonal decomposition of the covariance matrix $cov (W)$ of the above formula removes the correlation between the characteristics of the mass data of the Internet of Things, and enhances the dimensionality reduction processing effect of the mass data of the Internet of Things. The resulting matrix factorization is as follows: $\begin{matrix} (3) & cov (W) = R^{T} \otimes R \end{matrix}$

In formula (3), R is an orthogonal matrix, which is composed of eigenvectors in the covariance matrix, and $R^{T}$ represents a transposed orthogonal matrix.

Arrange the eigenvalues of the IoT mass data corresponding to each vector in descending order, and obtain the first few eigenvectors with the largest values. According to the obtained vector, the optimal subspace for the distribution of IoT mass data is determined, in which the projection operation of the initial IoT mass data is completed, and the new IoT mass data after dimensionality reduction processing is obtained.

Since the characteristic values of IoT mass data are in a state of continuous change, the time series of IoT mass data is established based on interval number theory [7,18] to better suppress the fuzzy uncertainty factors in IoT mass data. The fuzzy Internet of Things mass data attributes and the change state have a proportional function relationship, and the absolute change rate is used to describe the fuzzy attributes of the Internet of Things mass data.

Assuming that the mass IoT data corresponding to time t is $y_{t}$ , the absolute rate of change of the IoT mass data at t is calculated by the following expression: $\begin{matrix} (4) & U_{y} = | \frac{(y_{t} - y_{t - 1}) (y_{t - 1} - y_{t - 2})}{y_{t}} | \end{matrix}$

If the first-order difference of IoT mass data $y_{t}$ at time t is $P_{y}$ , then the fuzzy radius $O_{y}$ of IoT mass data is solved by the following formula: $\begin{matrix} (5) & O_{y} = \frac{\sum_{t}^{A} y_{t}}{U_{y} \times | P_{y} |} \end{matrix}$

In formula (5), A represents the total number of time series.

From this, the interval intuitionistic fuzzy number used to describe the $y_{t}$ attribute of the mass data of the Internet of Things is derived, as follows: $\begin{matrix} (6) & S_{y} = [y_{t} - O_{y}, y_{t} + O_{y}] \end{matrix}$

In the same way, the interval intuitionistic fuzzy numbers of all IoT mass data are obtained, and a time series $A (t) = {S_{1}, S_{2}, \dots, S_{y}}$ composed of intuitionistic fuzzy numbers in each interval is established, which reduces the difficulty and complexity of subsequent IoT mass data classification.

2.2. Construction of feature vector space model for mass data of Internet of things

After the mass data preprocessing of the Internet of things, the mass data feature items of the Internet of things are selected to build the feature vector space model of the mass data of the Internet of things, so as to form the feature vector of the mass data of the Internet of things.

A bit is the smallest data unit in a computer, and a bit is the characteristic item of the mass data of the Internet of Things. After the feature statistics of the IoT mass data set, if a certain IoT mass data feature item does not appear in each sample. Then the feature item is removed from the feature vector of the mass data of the Internet of Things, so as to reduce the dimension of the feature vector of the mass data of the Internet of Things.

Assume that the set of z IoT mass data to be classified is $X = {X_{1}, X_{2}, \dots, X_{z}}$ , and $c_{z}$ is the feature vector of the IoT mass data set $X_{z}$ . Each IoT mass data has v feature items, and $c_{z v}$ is the weight of the v feature item in the z IoT mass data. Then the eigenvector space model for building the mass data of the Internet of Things is: $\begin{matrix} (7) & B = \frac{1}{X_{z}} c_{z} \sum_{z = 1}^{v} c_{z v} \end{matrix}$

Use relevant data to describe the weight of feature items in the mass data of the Internet of Things, and use TF-IDF expression to express: $\begin{matrix} (8) & d_{i k} = X_{z α} \times β_{α} \end{matrix}$

In formula (8), $X_{z α}$ represents the number of times the IoT mass data feature item α appears in the IoT mass data set $X_{z}$ , and $β_{α}$ represents the inversely proportional data frequency of the IoT mass data feature item α. The most commonly used calculation formula is: $\begin{matrix} (9) & β_{α} = log (\frac{z}{H_{γ}} + 0.01) \end{matrix}$

In formula (9), $H_{γ}$ represents the amount of IoT mass data in which the IoT mass data feature item α appears in the IoT mass data set $X_{z}$ . Then, the equalization normalization process is performed [9,19], and the calculation formula is as follows: $\begin{matrix} (10) & F_{S} = \frac{\sqrt{X_{z α} \times log (\frac{z}{H_{γ}} + 0.01)}}{\sqrt{B \times \sum_{α = 1}^{v} X_{z α} \times log (\frac{z}{H_{γ}} + 0.01)}} \end{matrix}$

Based on the above process, the feature vector of the mass data of the Internet of Things is formed.

2.3. Classification of Internet of things mass data based on fuzzy clustering maximum tree algorithm

After the feature vector space model of IoT mass data is constructed, z IoT mass data sets $X_{z}$ are transformed into z v-dimensional sparse unit vectors. Using the fuzzy clustering maximum tree algorithm [14,15,17], the fuzzy clustering analysis is carried out on the mass data collection of the Internet of Things, and the classification of the mass data of the Internet of Things is realized.

The fuzzy similarity matrix for establishing the mass data set $X_{z}$ of the Internet of Things is expressed as: $\begin{matrix} (11) & G = {(g_{z x})}_{z \times z} \end{matrix}$

That is to say calibration, the key task is to determine the similarity coefficient that represents the similarity between the IoT mass data set $X_{z}$ and the IoT mass data set $X_{x}$ and is expressed as: $\begin{matrix} (12) & g_{z x} = sim (X_{z}, X_{x}) \end{matrix}$

For the determination of $g_{z x}$ , the commonly used method is the correlation coefficient method expressed as: $\begin{matrix} (13) & g_{z x} = \frac{| \sum_{α = 1}^{v} (X_{z α} - X_{z}) (X_{z α} - X_{x}) |}{\sqrt{\sum_{α = 1}^{v} {(X_{z α} - X_{z})}^{2}} \sqrt{\sum_{α = 1}^{v} {(X_{z α} - X_{x})}^{2}}} \end{matrix}$

In order to improve the accuracy of similarity measurement, a geometric arithmetic mean method is used for calculation: $\begin{matrix} (14) & g_{z x} = \frac{\sum_{α = 1}^{v} X_{z α} \sqrt{X_{z} \times X_{x}}}{\sum_{α = 1}^{v} X_{z α} \frac{X_{z} + X_{x}}{2}} \end{matrix}$

The connection between any two nodes $x_{z}$ and $x_{x}$ in the mass data set $X_{z}$ of the Internet of Things is expressed as an edge: $\begin{matrix} (15) & δ_{z x} = (z, x) \end{matrix}$

And the weight of the edge $δ_{z x}$ connected between two nodes in the mass data of the Internet of Things is: $\begin{matrix} (16) & ε_{z x} = ε (δ_{z x}) = g_{z x} \end{matrix}$

The resulting fuzzy graph is expressed as: $\begin{matrix} (17) & M = (K, L) \end{matrix}$

The construction steps of the fuzzy clustering maximum tree algorithm are as follows:

Step 1: Initialize the IoT mass data weight set ϵ, the IoT mass data node set θ, and the IoT mass data set ϑ connected between two nodes are all empty sets;

Step 2: Find the edge $δ_{z x}$ that connects two nodes in the IoT mass data with the largest weight in the fuzzy graph M;

Step 3: Put the weight $ε_{z x}$ on the edge $δ_{z x}$ connected between two nodes in the mass data of the Internet of Things into ϵ. Put a new node on the edge $δ_{z x}$ that connects two nodes in the mass data of the Internet of Things into θ. Put the edge $δ_{z x}$ connecting two nodes in the mass data of the Internet of things into ϑ. If $| θ | = z$ or $| ϑ | = z - 1$ , go to step 5; Otherwise, go to step 4.

Step 4: Check the weights of the edges connecting the two nodes in the IoT mass data composed of each node in θ and nodes other than θ, and find the connection between the two nodes in the IoT mass data with the largest weight edge $δ_{z x}$ , go to step 3;

Step 5: End, at this time, the edge connected between two nodes in the IoT mass data in ϑ constitutes the largest fuzzy clustering tree of M.

A certain threshold μ is selected to make a horizontal cut set for the fuzzy clustering maximum tree algorithm, and the edge connecting between two nodes in the IoT mass data less than μ in the fuzzy clustering maximum tree algorithm is disconnected. Then a forest containing several subtrees is obtained, in which each node of each subtree constitutes a class, and the number of trees is the corresponding number of classes. Through the above steps, the classification of mass data of the Internet of Things is realized.

3. Experimental analysis

In order to verify the effectiveness of the fast classification method of IoT mass data based on the fuzzy clustering maximum tree algorithm, this experiment is carried out under the Linux Centos operating system, the programming environment is Python ide, the code is written using the Numpy package, and the hardware environment is CPU Intel i5 2.8 Ghz, memory is 8 G.

3.1. Experimental data

The data used in the experiment comes from the Function dataset, and the properties of the simulated dataset include age, salary, vocation, elevel and other attributes. The mass data volume of IoT is selected as 5000 PB, and the mass data of IoT is quickly classified.

3.2. Experimental protocols and indicators

The classification recall, classification root mean square error and classification time were used as performance indicators for analysis. The method of reference [10], the method of reference [6] and the proposed method were used to compare the proposed method to verify the effectiveness of the proposed method.

Classification recall rate: classification recall rate refers to the proportion of the number of positive examples of correctly classified IOT mass data to the number of positive examples of actual IOT mass data. The higher the classification recall rate, the better the classification effect of the method’s IOT mass data. The calculation formula of classification recall rate is: $\begin{matrix} (18) & Recall = \frac{TP}{TP + FN} \end{matrix}$

In formula (18), $TP$ represents a real example and $FN$ represents a false counterexample.

Root mean square error of classification: the smaller the root mean square error of classification, the higher the classification accuracy of the method’s mass data of the Internet of things. The calculation formula is as follows: $\begin{matrix} (19) & RMSE = \sqrt{\frac{1}{ρ} \sum_{i = 1}^{ρ} {(σ_{i} - τ_{i})}^{2}} \end{matrix}$

In formula (19), ρ is the amount of classified IOT mass data, and $σ_{i}$ and $τ_{i}$ are the actual value and classification value of IOT mass data.

Classification time: classification time refers to the time consumed by different methods to complete the classification of mass data. The shorter the classification time, the higher the classification efficiency of the method.

3.3. Mass data classification recall rate of Internet of things

In order to verify the classification effect of the proposed method on the mass data of the Internet of things, the classification recall rate is taken as the evaluation index. Using reference [10] method, reference [6] method and the proposed method for comparison, we get the comparison results of the mass data classification recall rate of the Internet of things with different methods, as shown in Fig. 1.

Fig. 1.

Comparison of recall rates of IoT mass data classification by different methods.

Analysis of Fig. 1 shows that when the amount of IoT mass data is 5000 GB, the average IoT mass data classification recall rate of the method of reference [10] is 88.3%, the average IoT mass data classification recall rate of the method of reference [6] is 79.2%. The average IoT mass data classification recall rate of the proposed method is as high as 97.5%. It can be seen that the proposed method has a high recall rate of IoT mass data classification, indicating that the proposed method has a good IoT mass data classification effect.

3.4. Root mean square error of mass data classification in the Internet of things

On this basis, the accuracy of the proposed method is verified, and the root mean square error of classification is used as the evaluation index. By comparing the methods of reference [10], reference [6] and the proposed methods, the comparison results of root mean square error of mass data classification of the Internet of things with different methods are shown in Table 1.

Table 1
Comparison results of the root mean square error value of IoT mass data classification by different methods

IoT mass data volume/GB The proposed method The method of reference [10] The method of reference [6]

1000 0.023 0.052 0.075

2000 0.027 0.054 0.071

3000 0.038 0.043 0.068

4000 0.035 0.059 0.064

5000 0.029 0.061 0.070

Average value 0.030 0.054 0.069

IoT mass data volume/GB	The proposed method	The method of reference [10]	The method of reference [6]
1000	0.023	0.052	0.075
2000	0.027	0.054	0.071
3000	0.038	0.043	0.068
4000	0.035	0.059	0.064
5000	0.029	0.061	0.070
Average value	0.030	0.054	0.069

According to the data in Table 1, when the amount of IoT mass data reaches 5000 GB, the average root mean square error of the IoT mass data classification the method of reference [10] is 0.054, the average root mean square error of the IoT mass data classification the method of reference [6] is 0.069. The root mean square error value of the proposed method for the classification of mass IoT data is only 0.030. It can be seen that, compared with the method of reference [10] and the method of reference [6], the proposed method has a smaller root mean square error value for the classification of the Internet of Things mass data, indicating that the proposed method has a higher classification accuracy for the Internet of Things mass data.

3.5. Time comparison of IoT mass data classification

The proposed method is further validated for the classification time of IoT mass data. Comparing the method of reference [10] and the method of reference [6] with the proposed method, the comparison results of the classification time of the mass data of the Internet of Things of different methods are obtained as shown in Table 2.

Table 2
Comparison results of IoT mass data classification time by different methods

IoT mass data volume/GB The proposed method/ms The method of reference [10]/ms The method of reference [6]/ms

1000 2.3 5.9 9.8

2000 5.1 9.3 13.7

3000 8.6 13.5 16.8

4000 10.2 15.6 19.6

5000 12.3 19.5 24.7

IoT mass data volume/GB	The proposed method/ms	The method of reference [10]/ms	The method of reference [6]/ms
1000	2.3	5.9	9.8
2000	5.1	9.3	13.7
3000	8.6	13.5	16.8
4000	10.2	15.6	19.6
5000	12.3	19.5	24.7

According to the data in Table 2, with the increase in the amount of IoT mass data, the classification time of IoT mass data by different methods increases. When the IoT mass data volume is 5000 GB, the IoT mass data classification time of the method of reference [10] is 19.5 ms, the classification time of the Internet of Things mass data of the method of reference [6] is 24.7 ms, while the classification time of the Internet of Things mass data of the proposed method is only 12.3 ms. It can be seen that the proposed method can shorten the classification time of mass data in the Internet of Things.

4. Conclusion

This paper proposes a fast classification method for mass data in the Internet of Things based on the fuzzy clustering maximum tree algorithm. By preprocessing the mass data of the Internet of Things, select the characteristic items of the mass data of the Internet of things, and construct the feature vector space model of the mass data of the Internet of things. On this basis, using the fuzzy clustering maximum tree algorithm, fuzzy clustering analyzes the mass data collection of the Internet of Things, and realizes the classification of the mass data of the Internet of Things. The following conclusions are drawn from the experiments:

(1) The average IoT mass data classification recall rate of the proposed method is as high as 97.5%, which has a good IoT mass data classification effect.

(2) The root mean square error value of the proposed method for the classification of mass IoT data is only 0.030, which can effectively improve the classification accuracy of mass IoT data.

(3) The classification time of the IoT mass data of the proposed method is only 12.3 ms, which can shorten the IoT mass data classification time.

Footnotes

Acknowledgements

This paper is supported by Soft science Research Project of Henan Province ‘Research on the implementation of ‘three evaluations’ reform in Henan Province under the background of scientific and technological system reform’ (Grant no. 222400410238).

References

Du, Energy analysis of Internet of things data mining algorithm for smart green communication networks, Computer Communications 152(15) (2020), 223–231. doi:10.1016/j.comcom.2020.01.046.

Ismail and

Pham, Robust Markowitz mean-variance portfolio selection under ambiguous covariance matrix, Mathematical Finance 29(1) (2019), 174–207. doi:10.1111/mafi.12169.

Li,

Zhu,

Zhang and

Zhang, Design and implementation of intelligent traffic and big data mining system based on Internet of things, Journal of Intelligent and Fuzzy Systems 38(2) (2020), 1–9. doi:10.3233/JIFS-179971.

Lu,

Feng,

Chen,

Liu,

Lin and

Yan, Tensor robust principal component analysis with a new tensor nuclear norm, IEEE Transactions on Pattern Analysis and Machine Intelligence 42(4) (2020), 925–938. doi:10.1109/TPAMI.2019.2891760.

Lv,

Lou,

Li,

A.K.

Singh and

Song, Big data analytics for 6G-enabled massive Internet of things, IEEE Internet of Things Journal 8(7) (2021), 5350–5359. doi:10.1109/JIOT.2021.3056128.

Y.S.K.

Manjunath,

Zhao and

X.P.

Zhang, Time-distributed feature learning in network traffic classification for Internet of things, in: 2021 IEEE 7th World Forum on Internet of Things (WF-IoT), Vol. 29, IEEE, 2021, pp. 674–679. doi:10.1109/WF-IoT51360.2021.9595307.

Mi,

Liu,

Xiao,

Liu and

Yao, Interval multi-objectives optimization of electric wheel dump truck frame based on blind number theory, Applied Sciences 9(20) (2019), 4214–4223. doi:10.3390/app9204214.

S.M.

Mujeeb,

R.P.

Sam and

Madhavi, Energy- and trust-aware secure routing algorithm for big data classification using MapReduce framework in IoT networks, International Journal of Modeling, Simulation, and Scientific Computing 12(2) (2021), 215–221. doi:10.1142/S1793962321500100.

Park and

C.D.

Yoo, CNN-based learnable gammatone filterbank and equal-loudness normalization for environmental sound classification, IEEE Signal Processing Letters 27(2) (2020), 411–415. doi:10.1109/LSP.2020.2975422.

10.

Postol,

Diaz,

Simon and

Wicke, Time-series data analysis for classification of noisy and incomplete Internet-of-things datasets, in: 2019 18th IEEE International Conference on Machine Learning and Applications (ICMLA), Vol. 16, IEEE, 2020, pp. 193–202.

11.

Rghioui,

Lloret and

O.A.

Big, Data classification and Internet of things in healthcare, International Journal of E-Health and Medical Communications 11(2) (2020), 20–37. doi:10.4018/IJEHMC.2020040102.

12.

Rouani,

M.F.

Harkat,

Kouadri and

Mekhilef, Shading fault detection in a grid-connected PV system using vertices principal component analysis, 164(C) (2021), 1527–1539.

13.

M.A.

Sulaiman, Evaluating data mining classification methods performance in Internet of things applications, Journal of Soft Computing and Data Mining 1(2) (2020), 11–25.

14.

Tan,

Zhong,

Jiang,

Peng and

He, High-order fuzzy clustering algorithm based on multikernel mean shift, Neurocomputing 385(8) (2020), 63–79. doi:10.1016/j.neucom.2019.12.030.

15.

Tao,

Wang,

Chang and

Li, Density-sensitive fuzzy kernel maximum entropy clustering algorithm, Knowledge-Based Systems 166(2) (2019), 42–57. doi:10.1016/j.knosys.2018.12.007.

16.

Tsalera,

Papadakis and

Samarakou, Novel principal component analysis-based feature selection mechanism for classroom sound classification, Computational Intelligence 37(4) (2021), 1827–1843. doi:10.1111/coin.12468.

17.

Wang,

Wang and

Cui, Fuzzy clustering recognition algorithm of medical image with multi-resolution feature, Concurrency and Computation: Practice and Experience 32(1) (2020), e4886.1–e4886.12.

18.

Wang,

Xia,

Li and

Cao, Multi-criteria decision making method based on improved cosine similarity measure with interval neutrosophic sets, International Journal of Intelligent Computing and Cybernetics 12(3) (2019), 414–423. doi:10.1108/IJICC-05-2019-0047.

19.

Wu and

Li, Digital image steganography capacity normalization method based on DCT domain, Computer Simulation 38(8) (2021), 207–211.

20.

Zhang and

Zhang, A big data mining and blockchain-enabled security approach for agricultural based on Internet of things, Wireless Communications and Mobile Computing 2020(1) (2020), 1–8.

A fast classification method of mass data in Internet of things based on fuzzy clustering maximum tree algorithm

Abstract

Keywords

1. Introduction

2. Design of fast classification method for mass data in Internet of things

2.1. Mass data preprocessing of Internet of things

2.2. Construction of feature vector space model for mass data of Internet of things

2.3. Classification of Internet of things mass data based on fuzzy clustering maximum tree algorithm

3. Experimental analysis

3.1. Experimental data

3.2. Experimental protocols and indicators

3.3. Mass data classification recall rate of Internet of things

Table 2 Comparison results of IoT mass data classification time by different methods IoT mass data volume/GB The proposed method/ms The method of reference [10]/ms The method of reference [6]/ms 1000 2.3 5.9 9.8 2000 5.1 9.3 13.7 3000 8.6 13.5 16.8 4000 10.2 15.6 19.6 5000 12.3 19.5 24.7

Footnotes

Acknowledgements

References

Table 2
Comparison results of IoT mass data classification time by different methods

IoT mass data volume/GB The proposed method/ms The method of reference [10]/ms The method of reference [6]/ms

1000 2.3 5.9 9.8

2000 5.1 9.3 13.7

3000 8.6 13.5 16.8

4000 10.2 15.6 19.6

5000 12.3 19.5 24.7