Machine Learning Techniques and Analytics for Cloud Security. Группа авторов
Читать онлайн книгу.
hierarchical clustering is the second algorithm. It groups similar objects into groups (cluster). In this algorithm, it basically treats every observation as an individual cluster. After that, it iterates the following steps continuously:
(1) At first, consider the two clusters or groups that are closest together.
(2) Then, combine the two most similar clusters. Until all the clusters are combined together, this process continues [24].
The fuzzy c-means clustering is the last and third algorithm. This algorithm’s concept is very like to the k-means clustering. The algorithm is as follows:
(1) At first, identify clusters number.
(2) Then, randomly assign coefficients to each data point for being in the clusters.
(3) Until the algorithm has converged, repeats (1) and (2) step:(i) Compute centroid of each cluster or group.(ii) For every data point, compute the coefficient of being in the cluster.
2.3 Result
Result section consists of description of datasets, analysis of results, and validation of results.
Figure 2.1 Flowchart of the methodology.
2.3.1 Description of Datasets
Influenza sequences (glycan dataset) are taken from the National Centre for Biotechnology Information. At first, to perform searching operation, Basic Local Alignment Search Tool (BLAST) has been applied on H1N1 infected human datasets of Influenza A/447/08 at Oklahoma, Influenza A/1138/08 at Oklahoma, and Influenza A/447/08 at Oklahoma and on non-infected normal human of Influenza A/California/04/2009-4C. The dataset of H1N1 contains glycan data in Oklahoma City and the dataset of normal human contains glycan data in California City. The dataset consists of 442 different glycans and list of linkers are sp0, sp8, sp9, sp12, etc. Individual columns of the dataset represent the glycan numbers, glycan structure, the RFU, the STDEV value, and the SEM.
2.3.2 Analysis of Result
In this paper, unsupervised machine learning method like as k-means, hierarchical, and fuzzy c-means algorithm are shown to prove excellent classification performance and have been successfully applied in data analysis of H1N1 infected and non-infected datasets. At first, k-means clustering algorithm are applied on H1N1 infected dataset Influenza A/447/08 at Oklahoma, Influenza A/1138/08 at Oklahoma, and Influenza A/447/08 at Oklahoma and on non-infected dataset Influenza A/California/04/2009-4C that are shown in Figures 2.2 to 2.5. Same process will be repeated for hierarchical clustering algorithms that are shown in Figures 2.6 to 2.8. Fuzzy c-means has applied on above-mentioned datasets that are shown in Figures 2.9 to 2.11. After completing cluster analysis, we have collected those glycan structures where the value of RFU, STDEV, and SEM has been significantly changed from normal state to infected state.
Figure 2.2 K-means cluster analysis of Influenza A (H1N1) non-infected human.
Figure 2.3 K-means cluster analysis of Influenza A (H1N1) infected human.
Figure 2.4 K-means cluster analysis of Influenza A (H1N1) infected human.
Figure 2.5 K-means cluster analysis of Influenza A (H1N1) infected human.
2.3.3 Validation of Results
2.3.3.1 T-Test (Statistical Validation)
The t-test statistical validation has been applied for comparing the means of two samples (infected and normal), even if they have different number of glycans. The following steps are used to solve t-test validation:
a) List H1N1 infected datasets for sample 1.
b) List normal dataset for sample 2.
c) Record the number replicates (in the data set, n = 3) for sample (The number of replicates for sample1, i.e., n1 is 3, the number of replicates for sample2, i.e., n2 is 3).
d) Compute the mean of both n1 and n2 (x1’, x2’). [mean = total/n]
e) Compute the standard deviation (σ) for each sample (σ1, σ2). Where, σ2 = ∑d2/(n − 1)
f) Compute the variance that is the difference between the two means . Where
g) Compute σb (square root of ).
h) Compute the p value as follows:
Figure 2.6 Hierarchical cluster analysis of Influenza A (H1N1) infected human.
Figure 2.7 Hierarchical cluster analysis of Influenza A (H1N1) infected human.
Figure 2.8 Hierarchical cluster analysis of Influenza A (H1N1) infected human.
Figure 2.9 Fuzzy c-means cluster analysis of Influenza A (H1N1) infected human.
2.3.3.2 Statistical Validation