International Journal of Business Intelligence and Data Mining (17 papers in press)

Regular Issues

• Analysis of online marketing user participation preference attribute based on social network text mining  
  by Lu Zhang, Wanqing Chen, Hengzhi Nie 
  Abstract: In order to improve the accuracy of online marketing users' participation preference feature attribute recognition, an analysis method of online marketing users' participation preference feature attribute based on social network text mining is proposed. Firstly, the TF-IDF algorithm is used to calculate the weight value of keywords in the tag, and then the user portrait of the social network platform is constructed after sorting. Then, the collaborative filtering algorithm is used to determine the user's preference characteristics for products containing keywords, and the K-L feature compressor is used to extract the user's participation preference characteristics of online marketing. Finally, the online marketing user participation preference characteristic attributes are classified to realize the analysis of online marketing user participation preference characteristic attributes. The experimental results show that the accuracy of this method is always above 90% and the average time is 3.88s.
  Keywords: social networks; text mining; online marketing; preferential features; TF-IDF algorithm.
  DOI: 10.1504/IJBIDM.2025.10067625
   
  
• Improving brain MRI segmentation of multiple sclerosis using an advanced CNN approach  
  by V. Biksham , Sampath Korra, B. Pradeep Kumar , Salar Mohammad 
  Abstract: Multiple sclerosis (MS) can be detected early by looking for lesions in brain magnetic resonance imaging (MRI). Recently, unsupervised anomaly detection algorithms based on autoencoders were presented for the automatic identification of MS lesions. However, because these autoencoder-based approaches were created exclusively for 2D MRI pictures (e.g., 2D cross-sectional slices), they do not take use of the complete 3D information of MRI. In this research work, a novel 3D autoencoder-based methodological solution for detecting MS lesion volume in MRI is offered. We begin by defining a 3D convolutional neural network (CNN) for complete MRI volumes and then construct each encoder and decoder layer of the 3D autoencoder using 3D CNN. For optimal data reconstruction, we additionally include a skip link between the encoder and decoder layers. In the experimental results, we compare the 3D autoencoder-based method to the 2D autoencoder models using training datasets from the Human Connectome Project (HCP) and testing datasets from the Longitudinal MS Lesion Segmentation Challenge, and show that the proposed method outperforms the 2D autoencoder models by up to 20% in MS lesion prediction.
  Keywords: multiple sclerosis; brain MRI; image segmentation; CNN; chronic disease; healthcare.
  DOI: 10.1504/IJBIDM.2025.10068668
   
  
• AI advancements scary or hand holding for employees? A systematic literature review  
  by Remya Lathabhavan, Kottuvada M.S.V.D. Akshar 
  Abstract: The knowledge gained from a thorough literature analysis that was carried out to identify, categorise, and analyse recent developments in artificial intelligence (AI), its business applications, and its effects on the labour force is presented in this paper. Ninety-four papers are analysed and categorised as AI-related, business-related and domain-specific. AI developments and their applications in different functions of business and sectors of industry, and the possible impact on workforce are discussed. Robotic process automation, machine learning and natural language processing, along with their recent features that find use in business functions are presented. This study contributes to both technical and managerial literature. Future studies irrespective of their discipline can use this study as a roadmap from both technical and business perspectives. The paper also discusses the impact of AI on workforce in a futuristic and optimistic perspective. This study’s practical implications include illuminating the path towards individual self-evaluation and skill acquisition, organisational skill development and forecasting, and societal welfare policy framing.
  Keywords: AI advancements; AI and business applications; AI and workforce impact; AI and employees; systematic literature review.
  DOI: 10.1504/IJBIDM.2025.10068800
   
  

Special Issue on: Deep Learning Technology and Big Data Method for Business Intelligence and Management

• Learning from high-dimensional unlabelled data with outliers: a novel robust approach  
  by Abdul Wahid 
  Abstract: This paper investigates the problem of feature selection and classification under the presence of multivariate outliers in high-dimensional unlabeled data. The research question is how to identify outliers and deal with them in unsupervised learning to improve the clustering accuracy compared with the state-of-the-art non-robust feature selection techniques. For this purpose, a robust method is proposed by utilize the Mahalanobis distance for outlier identification based on the minimum regularized covariance determinants approach. Furthermore, a new weighting scheme based on Mahalanobis distance is developed for dealing with outlying data points. Finally, it is suggested to combine the proposed weight function and least squared loss function along with the graph and sparsity constraints for achieving the robustness. This new procedure is named Robust Self-Representation Sparse Reconstruction and Manifold Regularization (RSSRMR). The novel method is compared with previously proposed unsupervised feature select
  Keywords: clustering; high-dimensional data; feature selection; Mahalanobis distance; multivariate outliers.
  DOI: 10.1504/IJBIDM.2025.10066878
   
  
• Personalised recommendation method for smart library literature based on user behaviour feature perception  
  by Yina Liu 
  Abstract: To solve the problem that existing library literature recommendation methods cannot achieve diversity and personalisation, this study proposes a personalised recommendation method for smart library literature based on user behaviour feature perception technology. Firstly, based on network coding technology, the collection of user behaviour data for smart libraries is completed, and the collected data is reduced in dimensionality through information entropy to remove redundant features. Then, a user behaviour feature model is constructed through Bayesian networks to perceive and analyse user behaviour, obtain user behaviour features, and finally, based on the feature perception results, a collaborative filtering algorithm is used to complete personalised recommendation of literature materials in the smart library. The experimental results show that this method can fully utilise the behavioural characteristics of users, accurately understand their interests and needs, and provide more accurate literature recommendation results.
  Keywords: perception of user behaviour characteristics; smart library; literature materials; personalised recommendation; network coding; information entropy.
  DOI: 10.1504/IJBIDM.2025.10066987
   
  
• Method for mining students online English learning intention based on user portrait and big data  
  by Yanli Li, Lili Wang, Haitao Gao, Bin Zhang 
  Abstract: To overcome the problems of low recall, low accuracy, and long time in traditional methods, a new method for mining students' online English learning intention based on user portrait and big data is proposed. With the support of big data technology, the maximum mean difference algorithm is used to determine the distance between student online English learning data sample points, and the K-means algorithm is used to implement student online English learning data collection. The collected data is used to construct user personas, and the attention mechanism is used to extract students' online English learning characteristics. A student online English learning willingness mining model based on extreme learning machine network is established to obtain relevant mining results. Experimental tests have shown that the recall rate of the proposed method is always above 97.3%, the maximum mining accuracy is 98.1%, and the average mining time is 79.15ms.
  Keywords: user portrait; big data; students; online English; learning intention; maximum mean difference algorithm; attention mechanism; extreme learning machine.
  DOI: 10.1504/IJBIDM.2025.10066988
   
  
• Intelligent retrieval method for power grid dispatching information based on knowledge graph  
  by Baoyu Hou, Qichao Wang, Zhiguo Zhou 
  Abstract: To improve the retrieval efficiency of power grid dispatch information, the paper proposes an intelligent retrieval method based on knowledge graph. Firstly, after mining the terminology of power grid dispatch information, the entities and relationships in the power grid dispatch information are extracted to obtain a string of entity names in the terminology dictionary, achieving the design of the knowledge graph pattern layer for power grid dispatch information; Finally, the power grid dispatch information is embedded into a discrete Hamming space, and the nearest neighbour retrieval method is used in the embedded space to achieve intelligent retrieval of power grid dispatch information. The experimental results show that the intelligent retrieval accuracy of our method can reach 98.51%, the recall rate of our method can reach 98.32%, and the time consumption of our method is only 6.6 seconds. The retrieval efficiency of power grid dispatch information is relatively high.
  Keywords: knowledge graph; Hamming space; nearest neighbour retrieval; term dictionary tree.
  DOI: 10.1504/IJBIDM.2025.10066989
   
  
• A sentiment classification method for Weibo sensitive topic text based on multimodal features  
  by Manlin Li 
  Abstract: Due to the problem of reduced classification accuracy in traditional text sentiment classification methods, this paper proposes a Weibo sensitive topic text sentiment classification method based on multimodal features. Firstly, the bidirectional loop structure is introduced to improve the GRU model, and a BiGRU model is constructed for multimodal feature extraction and fusion of sensitive topics on Weibo. Secondly, by combining seed features, similar features, and residual features, a multimodal feature cluster is constructed to improve the accuracy of classification. Finally, the constructed multimodal feature clusters are input into the support vector machine model to complete sentiment classification of Weibo sensitive topic text. The experimental results show that compared with traditional methods, our method achieves higher accuracy in all emotion categories.
  Keywords: multimodal features; Weibo sensitive topics; text sentiment classification; BiGRU model; multimodal feature clusters.
  DOI: 10.1504/IJBIDM.2025.10066994
   
  
• An accurate and rapid pushing of marketing information based on multidimensional data mining  
  by Zhisheng Zhou, Bin Li 
  Abstract: In order to address the accuracy and recall issues in marketing information push, this study proposes a strategy based on multidimensional data mining to achieve accurate and efficient marketing information push. First of all, collect marketing information and build a push probability index system; Secondly, the analytic hierarchy process is used to calculate the weight of the marketing information push index; Finally, considering the product life cycle, data mining technology is used to obtain the stable and random purchase interest of active and inactive users, and marketing information is accurately and rapidly pushed through the above four dimensions. The research results show that after adopting this method, the accuracy of marketing information push reached 98.1%, the recall rate reached 96.9%, and user satisfaction also increased to 98.5%, significantly improving the overall effect of marketing information push and user satisfaction.
  Keywords: analytic hierarchy process; purchase interest; data mining techniques; multidimensional data mining; indicator weight.
  DOI: 10.1504/IJBIDM.2025.10066996
   
  
• A method for merging and classifying higher mathematics teaching resources based on density clustering algorithm  
  by Hejie Chang, Xing Lv 
  Abstract: To enhance the recall and accuracy of resource merging classification, this study introduces a merging classification technique rooted in density clustering algorithms. Initially, we gather data pertaining to higher mathematics teaching resources. Subsequently, we convert textual sentences into word-level representations, eliminating stop words and unnecessary high-frequency vocabulary. Leveraging LDA, we extract mathematical resource features, transforming words into computer- and model-recognisable vectorised forms. Next, we calculate the density and distance between samples to categorise them into distinct groups, employing density clustering algorithms for merging and classifying teaching resources. Experimental findings reveal that our method achieves a classification recall rate of 99.6% and an accuracy of 99.9%, thereby enhancing the quality and efficacy of higher mathematics education.
  Keywords: density clustering; merge and classify; advanced mathematics; teaching resources; resource allocation.
  DOI: 10.1504/IJBIDM.2025.10066997
   
  
• Research on engineering cost prediction based on GA-BP neural network  
  by Yan Wu, Sha Lan, Tingting Liu 
  Abstract: To improve the accuracy of engineering cost prediction and reduce prediction errors, an engineering cost prediction method based on GA-BP neural network is proposed in this paper. Comprehensive index system for engineering cost prediction is constructed, and qualitative indicators are discretized using the equal interval method. The qualitative indicators are transformed into quantitative indicators through scale assignment. The BP neural network error is obtained through gradient descent, and the GA algorithm is used to adjust the weights from the output layer to the hidden layer. Using the discretized qualitative indicators as input vectors and engineering cost as the output vector, a prediction model for engineering cost based on GA-BP neural network is built to obtain prediction results. Experimental results show that the proposed method has a prediction range of 2.41%, a residual mean range of 0.005~0.219, a recall rate fluctuating between 96.9% and 99.7%,and high prediction accuracy.
  Keywords: GA algorithm; BP neural network; engineering cost prediction; gradient descent.
  DOI: 10.1504/IJBIDM.2025.10066998
   
  
• Intelligent evaluation method for multimedia network public opinion decline period based on multi-divisional optimisation  
  by Xuefang Zhou 
  Abstract: In order to overcome the long data collection time, low accuracy in extracting features of public opinion decline, and low precision rate associated with traditional methods, a new intelligent evaluation method for multimedia network public opinion decline period based on multi-divisional optimization is proposed. A evaluation index system for intelligent evaluation of public opinion decline period is constructed, and index data is collected and processed. The multiple fractal dimensions of the index data are determined, and multi-divisional optimization is performed in conjunction with nonlinear support vector machines to extract features of public opinion decline. Public opinion decline period intelligent evaluation is achieved based on these features and the BiLSTM model. The experimental results show that the average data collection time of the proposed method is 0.72s, the average accuracy of feature extraction of public opinion decline is 97.66%, and the precision rate is consistently above 95%.
  Keywords: multi-divisional optimisation; multimedia network; public opinion decline period; intelligent evaluation; nonlinear support vector machine; BiLSTM model.
  DOI: 10.1504/IJBIDM.2025.10066999
   
  
• Data mining method for English classroom teaching quality based on hierarchical clustering  
  by Yue Zhang 
  Abstract: English classroom teaching involves multiple types of data, and effectively collecting and organising these data is a challenging task. Therefore, a hierarchical clustering based data mining method for English classroom teaching quality is proposed. Firstly, use dynamic layered distributed data collection algorithms to collect data; Secondly, use a moving average filter to smooth the data, transform the data through Fourier transform, and calculate the threshold for outliers based on normal distribution to achieve data outlier handling. Then, the recursive feature elimination method is used to perform feature selection on the data, and linear discriminant analysis is used to perform feature dimensionality reduction. Finally, use hierarchical clustering algorithm for data mining. The experimental results show that the recall rate of this method is high, the mean square error is low, the data storage space occupied is low, indicating that this method can effectively improve the effectiveness of teaching mining.
  Keywords: hierarchical clustering; data mining; recursive feature elimination; normal distribution.
  DOI: 10.1504/IJBIDM.2025.10067363
   
  
• A refined pushing method for financial product marketing data based on user interest mining  
  by Huijun Wang 
  Abstract: To improve the marketing effectiveness of financial products, the article designs a refined push method based on user interest mining. Divide user groups using K-means clustering algorithm and use density parameters to reflect user activity. Simultaneously, comprehensively explore users' interest and preferences in financial products. Build a marketing data model by calculating the transition probability of user browsing message categories. Finally, precise push is achieved by calculating the similarity between users, user interests, and candidate data. The experimental results show that after applying this method, the click through rate of financial products ranges from 93.58% to 96.69%, and the conversion rate of push results ranges from 0.88 to 0.95. The activity level of users participating in activities has always remained above 95%, verifying the effectiveness of this method.
  Keywords: Financial product marketing; Marketing data; User interests; Data push; K-means clustering; Category model.
  DOI: 10.1504/IJBIDM.2025.10067364
   
  
• Enhancing multiple document summarisation with DNETCNN and BCHOA techniques  
  by Mamatha Mandava, Surendra Reddy Vinta 
  Abstract: Multi-document summarising (MDS) is a helpful method for information aggregation that creates a clear and informative summary from a collection of papers linked to the same subject. Due to the significant number of information available online, it might be challenging to extract the needed information from an internet source these days. To generate the summary, we propose the binary chimp optimisation algorithm (BChOA) in this research. Several preprocessing techniques utilised to remove unwanted terms from the content. Then, for word embedding, FastText is used. The semantic and synthetic features are extracted using the DarkNet-53 and ConvNeXt methods. Using a darknet convolutional neural network (DNetCNN), the features derived from the syntactic and semantic features are concatenated. The Movie review dataset contains 2000 review files, and the BBC news dataset has 50 unique documents. Finally, the outcome demonstrates that our model compares to cutting-edge solutions in terms of semantics and syntactic structure.
  Keywords: multi-document summarisation; MDS; binary chimp optimisation algorithm; BChOA; ConvNeXt approach; darknet convolutional neural network; DNetCNN.
  DOI: 10.1504/IJBIDM.2025.10067365
   
  
• A precision marketing method for e-commerce considering the hidden behavioural characteristics of user online shopping  
  by Zheng Xu, Hengzhi Nie, Wanqing Chen 
  Abstract: In order to improve user order rate and achieve high marketing satisfaction, this article considers the hidden behaviour characteristics of user online shopping and designs an e-commerce precision marketing method. Firstly, use web crawler technology to collect and preprocess hidden behaviour data of e-commerce platform users during online shopping. Then, calculate the level of interest of e-commerce platform users in different labelled products during the online shopping process, and use natural language processing algorithms to identify the hidden behaviour characteristics of e-commerce platform users during online shopping. Based on the K-means clustering algorithm, perform fuzzy clustering on the hidden behaviour characteristics of online shopping. Finally, the Pearson similarity algorithm is used to calculate the similarity between feature data and target product data, and to construct an e-commerce platform’s online shopping product push matrix. Based on the ranking results of product push, precise e-commerce marketing is achieved. The experimental results show that using the proposed method, user satisfaction with product marketing recommendations remains above 85%, and user order rates remain above 90%. E-commerce marketing has high accuracy and good marketing effects.
  Keywords: e-commerce; electronic commerce; invisible behaviour; precision marketing; behavioural characteristics; fuzzy clustering.
  DOI: 10.1504/IJBIDM.2025.10067855
   
  

Special Issue on: Empowering Business Intelligence with AI Data Analytics and IoT for Efficient in Digital Era

• Presenting a model to reduce students' academic drop by using analytical comparison of machine learning algorithms in data mining (case study of Shahed University)  
  by Mozhdeh Salari, Reza Radfar, Mahdi Faghihi 
  Abstract: This research aims to find factors that predict undergraduate student educational performance. To achieve this goal, the study follows the CRISP-DM method. This study used various classification algorithms to predict the total GPA. The data used in this research are records of undergraduate students from 2012 in Shahed University. We used 1468 data records in data mining. We used the Rapidminer9.9 tool for modelling. This study also considers four feature selection techniques. This study used K-fold cross-validation to split the data. This study introduced the best model for predicting students' academic performance. In two-class modelling, we get better results and higher accuracy than four-class modelling. This research found the random forest algorithm best for predicting students performance. It achieved 94.17% accuracy with two classes. The random forest results show a higher chance of success in students with a higher 1st semester GPA.
  Keywords: student performance prediction; data mining; machine learning; data science applications in education.
  DOI: 10.1504/IJBIDM.2025.10067362