International Journal of Intelligent Information and Database Systems (4 papers in press)

Regular Issues

• Intelligent rainfall forecasting model: heuristic assisted adaptive deep temporal convolutional network with optimal feature selection  
  by Nishant Nilkanth Pachpor, B. Suresh Kumar, Prakash S. Prasad, Salim G. Shaikh 
  Abstract: A deep learning technology is adopted to predict seasonal rainfall efficiently. Various rainfall data are collected from the internet. A deep feature extraction is done by autoencoder. Further, the deep extracted features are provided to the optimal feature selection phase, where the weights are optimised by utilising the developed modified attack power-based sail fish-hybrid leader optimisation (MAP-SFHLO). Then, the selected optimal features are provided as input to the prediction stage, and the prediction is done using the enhanced atrous-based adaptive deep temporal convolutional network (EA-ADTCN) along with the aid of the developed MAP-SFHLO algorithm to offer an effective prediction rate as the final outcome. Throughout the analysis, the performance of the developed model shows 5.2% and 6.0% regarding MAE and RMSE metrics. Thus, the suggested system performs more accurately in terms of accuracy rate in predicting rainfall than conventional techniques.
  Keywords: rainfall forecasting model; autoencoder-based deep feature extraction; optimal feature selection; modified attack power-based sail fish-hybrid leader optimisation; enhanced atrous based adaptive deep temporal convolutional network.
  
  
• Constructing a Chinese-Vietnamese bilingual corpus from subtitle websites  
  by Phuc-Nghi Nguyen, Phuoc Tran 
  Abstract: In this work, we introduce a method of constructing a Chinese-Vietnamese bilingual corpus on subtitle resources. The corpus construction process involved careful curation and preprocessing of the chosen subtitle data to ensure its suitability for training and evaluating machine translation models. We applied rigorous quality control measures to enhance the reliability and relevance of the collected corpus by systematically eliminating entries that did not meet a predetermined level of correctness. We use the two robust neural machine translation models to experiment on the collected corpus. The experimental results show that the highest BLEU score of the collected corpus is 22.0, much higher than the OpenSubtitles 2016 corpus one of the most popular subtitle corpus today. By curating a specialised corpus, we aim to contribute valuable resources to the field of machine translation, fostering advancements in the understanding and improvement of translation quality between Chinese and Vietnamese.
  Keywords: Chinese-Vietnamese bilingual corpus; machine translation; Netflix.
  
  
• Incorporating label-wise thresholding and class-imbalanced strategy into binary relevance for online multi-label classification  
  by Kunyong Hu, Tingting Zhai 
  Abstract: Binary relevance (BR) is widely used to solve multi-label classification problems. Typically, all binary classifiers in BR use a shared global fixed threshold to convert predicted values to binary classification results. However, some studies disclosed that tuning a separate threshold per label is better than a fixed global threshold. Given this discovery, in this paper, we adaptively train a thresholding model for the scoring model of each binary classifier in BR. By solving an online convex optimisation problem that minimises a logistic loss function, both models can be updated simultaneously. Furthermore, each binary classifier may suffer from the class-imbalance problem. To this end, we design three cost-sensitive strategies to adjust the misclassification cost of relevant and irrelevant labels for each binary classifier. An efficient closed-form update can be obtained by solving our formulated problem. Extensive experiments on multiple datasets demonstrate that our methods outperform other state-of-the-art methods.
  Keywords: online multi-label classification; binary relevance; adaptive label thresholding; class-imbalance; cost-sensitive learning.
  
  
• An extended recommendation system by applying aspect-based sentiment analysis method on customers’ review and experience  
  by Thanh Ho, An-Dinh Van, Trinh Tran Thi Kieu, Anh Nguyen Thi Linh, Thao Huynh Nhi Thanh, Hieu Tran Nguyen Ngoc 
  Abstract: The hotel industry heavily relies on customer reviews and recommendations to attract guests and ensure customer satisfaction. However, the abundance of online reviews poses a challenge for users seeking relevant and reliable information for their hotel selection process. This study proposes a novel approach that leverages aspect-based sentiment analysis (ABSA) to assess the alignment between hotel sentiment scores and user preferences at the aspect level. By experimenting on 62,887 customer feedback in hospitality industry, two approaches are developed and presented: 1 calculating aspect-based sentiment hotel profiles using Wu-Palmer similarity, predefined index terms, and PyABSA; 2) extracting user preferences through term frequency (TF). The recommendation results are obtained by aligning hotel sentiment scores with user preferences using a value-focused approach (scalar product). Our evaluation, based on the Mean Reciprocal Rank (MRR) metric, reveals that, the most suitable hotel is within the top 5 recommended items, with an MRR value of 0.2849.
  Keywords: aspect-based sentiment analysis; ABSA; recommendation system; context-awareness; preference extraction; customer experience; personalised recommendations.