International Journal of Intelligent Information and Database Systems (7 papers in press)

Regular Issues

• Hybrid jellyfish search sine cosine optimisation-based deep learning for big data classification using MapReduce framework on epileptic seizure data  
  by Jamunadevi Chandrasekar, Arul Ponnusamy 
  Abstract: The increase in the amount of big data with the technical advances makes the traditional software tools face difficulties and unable to handle them. In the medical field, big data technologies require new frameworks to leverage them. This paper proposes a novel big data classification using a MapReduce framework on epileptic seizure data is proposed. Here, the big data classification is accomplished in a MapReduce framework, wherein the mapper phase is applied with the data partitioned by deep embedded clustering (DEC). The classification is carried out in the reducer phase, where a deep long short-term memory (DLSTM) trained using the jellyfish search sine cosine (JSCS) algorithm is used for epileptic seizure detection based on the salient features determined from the EEG data. The JSCS-DLSTM is investigated for its efficiency based on accuracy, specificity, and sensitivity and is found to record superior values of 0.915, 0.927, and 0.919, respectively.
  Keywords: deep embedded clustering; DEC; jellyfish search sine cosine algorithm; JSCS; deep long short-term memory; big data classification; epileptic seizure detection.
  
  
• Sentiment classification method of fresh agricultural product reviews based on semantic and emotional optimisation  
  by Yindong Dong, Yu Zhang, Guodong Wu, Xia Chen, Lijing Tu, Guohua Fan 
  Abstract: The complexity of semantics and structure in fresh agricultural product reviews can lead to sparsity in the distribution of text sentiment information. To address this issue, a sentiment classification model (Electra XLNet BIGRU multi-head attention - EXBMA) for fresh agricultural product reviews is proposed, which combines semantic and sentiment information optimisation. Firstly, the Electra model and XLNet model were used to obtain word level and sentence level information of fresh agricultural product reviews, respectively, and the semantic features obtained were fused using (multimodal compact bilinear - MCB) algorithm. Secondly, TextRank is used to extract keywords to construct an emotional key dictionary and combined with a multi-head attention mechanism to enhance emotional attention. Finally, to enhance the contextual representation, the BIGRU is used to learn contextual information to improve the classification performance. The experimental results indicate that the EXBMA can better achieve collaborative optimisation of semantic and emotional information, and performs better than other existing classification models in emotional classification of fresh agricultural product reviews.
  Keywords: fresh agricultural products; emotional classification; semantic optimisation; sentiment lexicon; attention mechanism.
  
  
• Hybrid brave-hunting optimisation for heart disease detection model with SVM coupled deep CNN  
  by Pravin M. Tambe, Manish Shrivastava 
  Abstract: This research proposes a novel hybrid optimisation method called brave-hunting optimisation (BHO) inspired by lion optimisation (LO) and coyote optimisation (CO). The BHO algorithm is employed to fine-tune the SVM parameters, enhancing its classification performance. Simultaneously, a deep CNN model extracts complex and informative features from medical data. The combined approach capitalises on the strengths of both optimisation techniques to create a robust and accurate model for heart disease diagnosis. The performance evaluation of our model is conducted using comprehensive metrics, which achieve an accuracy of 94.89%, an F1-score of 94.48%, a precision of 94.48%, and a recall of 94.58% for a 90 TP. In the context of a ten k-fold evaluation, achieved 94.78% accuracy, 94.36% F1-score, 94.55% precision, and 94.13% recall.
  Keywords: cardiovascular disease detection; support vector machines; deep convolutional neural network; DCNN; brave-hunting optimisation; BHO; early prediction.
  
  

Special Issue on: Knowledge Extraction and Mining to Enhance Intelligent Information Systems

• MS-ConvNeXt: a deep-learning method for tomato leaf diseases identification  
  by Yunchao Li 
  Abstract: Existing deep learning methods for tomato leaf disease identification are challenged by the multi-scale disease regions and complex backgrounds in tomato leaf images. A network for tomato leaf disease is proposed. In the proposed network, a cross-channel-and-spatial attention mechanism is first introduced in the ConvNeXt block (called A-ConvNeXt block) to avoid interference of invalid features from the complex backgrounds. Then, a multiscale feature mechanism is integrated into the backbone constructed by the A-ConvNeXt block to extract features across multiscale diseases. The fine multiscale and silence features are extracted to address the limitations on tomato leaf diseases. Experimental results on laboratory and natural datasets show that the identification accuracy reached 95.67%, which outperformed many other existing networks in comparison experiments. The proposed network may effectively improve tomato leaf disease identification and provide decision-making information for practical applications in modern agriculture.
  Keywords: tomato leaf disease identification; attention mechanism; multiscale feature mechanism; deep learning.
  
  

Special Issue on: Multi-modal Information Learning and Analytics on Data Integration

• Design and optimisation of electrical information collection and transmission system based on multimodal information analysis  
  by Jinyin Peng, Xiangjin Zhu 
  Abstract: In order to improve the comprehensiveness of power information collection and the flexibility of transmission systems, this article combines multimodal information analysis to conduct in-depth research on system design and optimisation from the perspective of software and hardware structure, and tests it from four aspects: data collection quality, data transmission efficiency, accuracy, and system security. The results show that in terms of data transmission accuracy, the average power data transmission accuracy test result of the system in this article reaches about 92.93%; the average test result of AC simulation transmission accuracy reaches about 93.11%; the average accuracy test result of working condition data transmission reaches about 91.68%. From the experimental results, it can be seen that under traditional finite network technology, the transmission accuracy test results of the electrical energy data, AC analogue quantity, and operating condition data of the five test nodes are about 83.91%.
  Keywords: electrical information; multimodal information analysis; data collection and transmission; system design and optimisation; power monitoring.
  
  
• Optimising path planning and obstacle avoidance algorithm for electrical robots using multimodal information learning techniques  
  by Yang Qiu, Bo Zhou, Lingxiao Chen 
  Abstract: In response to the problems of poor adaptability to complex environments, low success rate of obstacle avoidance, and low accuracy of path planning in traditional path planning and obstacle avoidance algorithms, this paper uses multimodal information learning technology to optimise the path planning and obstacle avoidance algorithms of electric robots. Compared with traditional obstacle avoidance algorithms, optimising obstacle avoidance algorithms using deep learning techniques in multimodal information learning and constructing obstacle avoidance algorithms based on vision and dynamic programming can effectively improve the success rate of obstacle avoidance for electric robots. The average pathfinding time of the six groups studied in this article is 58.59 seconds, which is 4.94 seconds and 3.21 seconds lower than the average values of the ant algorithm and A * algorithm, respectively; in a dynamic obstacle environment, the obstacle avoidance success rate of the algorithm studied in this paper is 96.67%.
  Keywords: obstacle avoidance algorithm; path planning; multimodal information learning technology; electrified robot; Q-learning algorithm; ant colony optimisation; ACO.
  
  
• Application of differential privacy technology in multi-modal data sharing  
  by Zhihai Lu, Bin Wang, Nuanqing Ouyang 
  Abstract: The advent of data-driven technologies and artificial intelligence (AI) has led to an increasing demand for the sharing and analysing sensitive information. However, the paramount concern of preserving individual privacy poses a significant challenge. Hence, an algorithm named differential privacy in data sharing for AI (PrivShareAI) has been utilised. The objective is to enable secure and privacy-preserving data sharing in AI by implementing differential privacy measures and maintaining a balance between utility and privacy. The data-sharing paradigm uses sensitivity limits, noise-enhanced queries, and a universal, secure architecture enabled by a trusted server to encourage shared learning while maintaining maximum privacy. The proposed models efficiency is evaluated with baseline comparison studies with the following metrics: privacy guarantee, accuracy on varying parameters, privacy-utility trade-off, and privacy loss and accuracy measure.
  Keywords: artificial intelligence; differential privacy; data sharing; accuracy; Gaussian noise; gradients; privacy guarantee.