International Journal of Innovation and Sustainable Development (5 papers in press)

Special Issue on: OA Interdisciplinary Research of Energy Application, Governance, and Policy for Sustainability

• Open Access

  Machine learning model for dynamic cost assessment of power cable tunnel construction ( Free Full-text Access )

  by Ming Fang, Yifeng Lai, Handong Lu, Chuang Wang 
  Abstract: In order to improve the accuracy and adaptability of project cost evaluation, a dynamic cost evaluation model based on gradient boosting decision tree (GBDT) algorithm is proposed. In this study, dynamic evaluation refers to the updating of input variables based on stages and the rolling prediction of construction cost with the change of project information during the construction process. Data of 100 cable tunnel projects were collected and preprocessed, including tunnel length, buried depth, cross-sectional area, geological type, material price, construction period and construction cost. The results show that the model has better prediction performance when the decision tree depth is 5, the learning rate is 0.01 and the sub-sampling rate is 0.8. The proposed method can provide auxiliary support for the preliminary cost estimation, cost correction and cost early warning of cable tunnel projects.
  Keywords: cable tunnel; construction cost; dynamic assessment; machine learning model; cost composition.
  DOI: 10.1504/IJISD.2026.10080828
   
  

Special Issue on: OA Sustainable Education in the Age of Artificial Intelligence (AI)

• Open Access

  Qualitative study on teacher support behaviour in outdoor autonomous games in kindergartens ( Free Full-text Access )

  by Juan Shu 
  Abstract: This study aims to address the problem of teachers support behaviour in outdoor autonomous games in kindergartens being easily influenced by situational complexity and instantaneous interaction, resulting in structural misalignment. Qualitative research methods were used to collect 198 materials through non participatory observation, semi-structured interviews, and teacher reflection logs. 24 initial concepts were encoded and extracted, and clustered into four core categories: emotional, social, cognitive, and instrumental (reliability 0.93). A teacher support behaviour structure framework containing four dimensions and 10 indicators was constructed. The results show that cognitive support needs are the most prominent, while the actual level of emotional and social support provided is significantly lower than the needs. Based on this, we propose ways to enhance the effectiveness of support: building an accepting psychological environment, promoting deep interaction, improving observation and guidance abilities, and optimising the environment and rule mechanisms.
  Keywords: kindergarten; outdoor independent games; teacher support behaviour; qualitative analysis; situational interaction.
  DOI: 10.1504/IJISD.2026.10080430
   
  

• Open Access

  Study on adaptive learning path recommendation method based on multi-task learning model ( Free Full-text Access )

  by Ningxian Fu, Xiang Li 
  Abstract: To address the issues of single-dimensional learner state modelling and insufficient dynamic adaptability in existing learning path recommendation methods, this paper proposes an adaptive recommendation approach integrating multi-task learning and reinforcement learning. Multi-source data including learning logs, test performance, and physiological signals are collected. A multi-task learning framework with adaptive task weighting is constructed to simultaneously perform knowledge tracing, cognitive diagnosis, and emotion recognition, enabling multi-dimensional learner state modelling. The resulting state vectors are then introduced into a reinforcement learning environment, where a deep Q-network dynamically plans learning paths to optimise long term learning gains. Experimental results show that the proposed method achieves a recommendation accuracy of over 95%, an average knowledge improvement of 38.9%, and a recommendation time within 11 s, demonstrating its significant advantages in enhancing learning outcomes and system efficiency.
  Keywords: learning path; multi task learning model; adaptive recommendation; deep Q-network.
  DOI: 10.1504/IJISD.2026.10080431
   
  

• Open Access

  GRA-LSTM for intelligent evaluation of blended teaching effectiveness in higher education ( Free Full-text Access )

  by Huaxi Chen, Chaoyong Jia, Lei Zhang 
  Abstract: As a key direction in higher education digital transformation, blended learning suffers from issues like narrow evaluation dimensions, shallow data use, and untimely feedback. To address these, an intelligent evaluation model using GRA-LSTM is proposed. Firstly, establish a three-dimensional evaluation index system. This system covers teaching preparation, teaching process, and teaching effectiveness. Secondly, singular value decomposition interpolation and stacked denoising autoencoder are used to preprocess multi-source heterogeneous teaching data. Finally, key influencing factors are selected through grey relational analysis. These factors are input into the long short-term memory network to capture long-term dependencies and achieve dynamic modeling and prediction of teaching effectiveness. The results showed that the GRA-LSTM model achieved a prediction accuracy of 90% on the test set, with an average absolute error of less than 0.2.
  Keywords: blended teaching; teaching effectiveness evaluation; grey relational analysis; long short-term memory network; higher education.
  DOI: 10.1504/IJISD.2026.10080855
   
  

Regular Issues

• Assessing green innovation efficiency and emissions reductions in G20 economies: a data envelopment analysis and panel regression approach  
  by Maria Letiția Andronic, Gheorghița Dincă, Dana Adriana Lupșa-Tătaru 
  Abstract: This study investigates green innovation efficiency (GIE) and carbon dioxide emissions among the Group of Twenty (G20) economies over the period 1997-2019. Eco-innovation performance is evaluated using data envelopment analysis (DEA), incorporating both technical efficiency scores and the Malmquist productivity index (MPI). The results identify a leading benchmark economy distinguished by a robust implementation of green technological innovation and environmental protection strategies, as reflected in the highest total factor productivity (TFP) and catch-up effect within the sample. Moreover, a panel data regression model is employed to assess the influence of technological eco-innovation, green finance, and environmental and social factors on reducing carbon dioxide emissions. Half of the selected variables exhibit a beneficial effect on mitigating undesirable emissions, with one member state consistently emerging as a top performer in environmental sustainability. Overall, the findings highlight the importance of enhancing international cooperation on green technologies and promoting the diffusion of best practices across the G20.
  Keywords: GIE; green innovation efficiency; carbon dioxide emissions; sustainable development; DEA; data envelopment analysis; panel data regression; G20.
  DOI: 10.1504/IJISD.2026.10078890