International Journal of Data Mining and Bioinformatics (8 papers in press)

Regular Issues

• A Data Mining Method for Biomedical Literature Based on Association Rules Algorithm  
  by Xiaofeng Shi, Yaohong Zhao, Haijuan Du 
  Abstract: There are problems in the process of biomedical literature data mining, such as high data noise, low mining accuracy, and long-time consumption. Therefore, a biomedical literature data mining method based on association rule algorithm was designed. First, set up the extraction process of biomedical literature data, introduce the factor graph decomposition global extraction function, and establish a probabilistic database to speed up the extraction. Secondly, wavelet transform is used to denoise the data, improve the effectiveness of the extracted data, and classify it based on its importance. Finally, by setting association rules for biomedical literature data mining and introducing pre pruning methods on this basis, the time cost of calculating support is reduced, mining efficiency is improved, and combining confidence and dependency, a biomedical literature data mining model based on association rules is constructed to achieve the final mining. The results show that this method improves the accuracy of literature mining, reaching 99%, and effectively reduces the mining time, with a maximum time consumption of 1.7 seconds. It has strong application performance.
  Keywords: association rules; biomedical literature; data mining; wavelet transform; vector space; classification basis.
  DOI: 10.1504/IJDMB.2023.10057445
   
  
• Research on Human Health Status Recognition Based on Association Algorithm  
  by Taiping Jiang, Zhibing Wang, Lei Huang 
  Abstract: A human health status recognition method based on association algorithm is proposed to address the problems of low recognition accuracy, low correlation of health data collection, and long recognition time in existing human health status recognition methods. Firstly, a temperature sensor is used to collect human body temperature data. Secondly, the photoelectric capacitance method is used to collect heart rate and blood oxygen data. Once again, by setting the 3D coordinate system of human bone points and using the depth image coordinate system to determine the true distance of bone points, the collection of human bone related data is achieved. Finally, association algorithms are used to analyse the relationship between human health status data, construct a human health status recognition function, and complete the recognition of health status. The test results show that the accuracy of the proposed method for identifying human health status remains around 99%.
  Keywords: association algorithm; identification of human health status; body temperature data; bone data; identification function.
  DOI: 10.1504/IJDMB.2023.10058130
   
  
• Low Resolution Face Recognition Algorithm Based on MB-LBP  
  by Bin Fang 
  Abstract: Due to the low accuracy, poor stability, and long time consumption of current face recognition methods, a low resolution face recognition algorithm based on MB-LBP is proposed. Firstly, the facial edge image is processed through binarisation, followed by scale normalisation to accurately locate the face and the final cropped facial image; then, segmented linear transformation is used for image enhancement processing; finally, MB-LBP is used to extract features, and the Euclidean distance and cosine angle between the extracted feature vectors and the feature vectors extracted from the face database are calculated to achieve dual matching of facial images and achieve face recognition. The results show that the quality of the results obtained by this algorithm is good, with peak signal-to-noise ratio and recognition accuracy of 160 dB and 100%, variance of 0.01, and recognition time of 1.8 s, indicating that the algorithm proposed in this paper has reliable application performance.
  Keywords: MB-LBP; low resolution; face recognition; binarisation; piecewise linear transformation.
  DOI: 10.1504/IJDMB.2023.10058131
   
  
• Data mining based integration method of infant critical and critical information in modern hospital  
  by Juan Xiao, Jina Zhang, Xiaoli Liu 
  Abstract: In this paper, a modern hospital infant emergency and critical information integration method based on data mining is designed. First of all, analyse the data types of children’s critical information in modern hospitals; then, metadata is extracted through mapping relationship; finally, the data missing value is filled in by the mean filling method, and the support and correlation of the data are calculated by the association rule algorithm, and the information integration model is constructed to realise the information data integration. The test results show that the error of the proposed method for the integration of children’s critical and critical information in modern hospitals is always lower than 0.3%, the throughput is always above 75 Mbps, and the maximum integration time is only 2.12 s, which has good practical application performance.
  Keywords: data mining; modern hospitals; children are in critical condition; information integration; metadata; variable linear method.
  DOI: 10.1504/IJDMB.2023.10058132
   
  
• Fast retrieval method of biomedical literature based on feature mining  
  by Duo Long, Yunxin Long, Fahuan Xie, Ping Yu, Hui Yan 
  Abstract: In order to solve the problems of large error, low accuracy of feature mining and long time consuming in traditional literature retrieval methods, this paper designs a fast retrieval method of biomedical literature based on feature mining. First, simulate the document collection space, and collect documents according to the data centroid and probability density function. Secondly, the location of similar data is marked by mutual information method, and the hidden information of literature data is extracted after reducing the imbalance of dataset. Then, the Pearson correlation coefficient of the literature data is calculated and the key features of the literature are mined. Finally, calculate the expected loss risk of literature data, design a fast retrieval algorithm for biomedical literature, and realise fast retrieval. The test results show that this method can reduce the retrieval error, improve the accuracy of document feature mining, and the retrieval time is shorter.
  Keywords: feature mining; biomedical literature; quick search; probability density function; cost matrix.
  DOI: 10.1504/IJDMB.2023.10058133
   
  
• Research on Cloud Storage Biological Data De duplication Method Based on Simhash Algorithm  
  by Haijuan Du 
  Abstract: Aiming at the problems of high duplication error, low data similarity accuracy and poor throughput in cloud storage biological data de duplication, a cloud storage biological data de duplication method based on Simhash algorithm is designed. First, analyse the cloud storage mode, characteristics and advantages of biological data, and determine the distribution rule of biological data in cloud storage; then, K-nearest neighbour algorithm and Bayesian algorithm are used to extract the features of cloud storage biological data; finally, Simhash algorithm is introduced to map the data into digital signatures that are longer than specialty to the maximum extent, set digital signatures of different dimensions after cloud storage biological data mapping, determine the similarity of biological data signature bit values, and complete duplication removal. The results show that the proposed method has lower error and is feasible.
  Keywords: Simhash algorithm; cloud storage; biological data; de duplication method; K-nearest neighbour algorithm; Bayesian algorithm; digital signature.
  DOI: 10.1504/IJDMB.2023.10058134
   
  
• Identification of disease-related miRNAs based on Weighted K-Nearest Known Neighbors and Inductive Matrix Completion  
  by Ahmet Toprak 
  Abstract: miRNAs, a subtype of non-coding RNAs, have a length of about 18-22 nucleotides. Studies have shown that miRNAs play an important role in the initiation and progression of many human diseases, mainly cancer types. For this reason, it is very significant to know the miRNAs associated with diseases. However, experimental studies to determine these relationships are a very expensive and time-consuming process. In this study, we propose a calculation method based on nearest known neighbors and matrix completion. ROC curves of our suggested method were plotted using two commonly used cross-validation techniques such as 5-fold and LOOCV, and also AUC values were calculated in both validation techniques. Moreover, we carried out case studies on breast cancer, lung cancer, and lymphoma in our method. We validated the results from experimentally validated databases. As a result, our proposed method can be used with confidence to identify possible miRNA-disease associations.
  Keywords: ncRNA; miRNA; disease; cancer; miRNA-disease associations.
  DOI: 10.1504/IJDMB.2023.10058135
   
  
• Diagnosis of Parkinson’s disease genes using LSTM and MLP based multi-feature extraction methods  
  by Priya Arora, Ashutosh Mishra, Avleen Malhi 
  Abstract: Disease gene identification using computational methods is one of the most challenging issues to improve the treatment and diagnosis of Parkinson’s disease (PD). Various intelligent computing techniques have been introduced to predict disease associated genes but the major difference among these approaches is in the data type to be used to create a feature vector. In this paper, deep learning methods such as multi-layer perceptron (MLP) and long short-term memory (LSTM) are adopted to identify genes that are responsible for Parkinson’s disease. The proposed method has been optimised on the basis of: 1) amino acid’s physicochemical properties to construct a feature vector; 2) feature extraction method to reduce the effect of noise and to speed up the process; 3) the genes prediction is done by employing deep learning methods. Compared with different type of datasets and other classifiers, the proposed method improves the prediction performance of neurodegenerative diseases. The experimental results indicate that the proposed deep learning approach outperforms the existing gene identification methods with higher recall, precision and F-score of 88.2, 84.5 and 85.0, respectively. The results of proposed system indicate the efficiency and accuracy for Parkinson’s disease gene identification and classification.
  Keywords: disease gene; deep learning models; feature extraction; physicochemical properties of amino acid; Parkinson’s disease.
  DOI: 10.1504/IJDMB.2023.10058573