Jiarui Zhou
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Jiarui Zhou has a total of 2 co-authors. Here are the top co-authors by number of shared publications:
- Chengxiao Dai
Faculty of Information and Communication Technology, Universiti Tunku Abdul Rahman, Malaysia
1 shared publication - Yanying Cheng
Faculty of Humanities and Social Sciences, Daugavpils University, LV-5401 Daugavpils, Latvia
1 shared publication
Latest Publications
- Open Access
- 24 May 2024
- Published by Association for Computing Machinery (ACM)
- p. 499-503
From the Perspective of Explainable Machine Learning: A Student Feature Selection Strategy Based on the Geometric Mean of Feature Importance and Robustness
Educational data mining is a growing field in education, but there is no consensus on the interpretability of data mining algorithms. We propose a student feature selection strategy based on the random forest algorithm that scales the gap between two features and adopts a geometric mean method to balance the contribution of performance importance and robustness, which is useful for educators and academics Researchers are vital. We collected a dataset of academic performance and characteristics of students from Guangdong Vocational College of Finance and Trade for experiments. Experiments show that our feature selection strategy is highly effective and produces relatively great results in both random forest and linear regression algorithms. Additionally, our student feature selection strategy provides valuable insights into the field of educational data mining by offering a clearer understanding of the factors influencing students ' academic performance.
- 24 May 2024
- Published by Association for Computing Machinery (ACM)
- p. 499-503
Open AccessFrom the Perspective of Explainable Machine Learning: A Student Feature Selection Strategy Based on the Geometric Mean of Feature Importance and Robustness
Educational data mining is a growing field in education, but there is no consensus on the interpretability of data mining algorithms. We propose a student feature selection strategy based on the random forest algorithm that scales the gap between two features and adopts a geometric mean method to balance the contribution of performance importance and robustness, which is useful for educators and academics Researchers are vital. We collected a dataset of academic performance and characteristics of students from Guangdong Vocational College of Finance and Trade for experiments. Experiments show that our feature selection strategy is highly effective and produces relatively great results in both random forest and linear regression algorithms. Additionally, our student feature selection strategy provides valuable insights into the field of educational data mining by offering a clearer understanding of the factors influencing students ' academic performance.
- Open Access
- 26 April 2024
- Published by Association for Computing Machinery (ACM)
- p. 59-65
Research and application of gaze direction algorithm based on head movement and eye movement data fusion
Visibility could potentially signify attentiveness. Estimating the line of sight is crucial in numerous fields, including computer vision, human-computer interaction, and psychology. Estimating the line of sight is crucial in numerous fields, including computer vision, human-computer interaction, and psychology. Line of sight estimation is difficult to achieve due to changes in head position and advancements in precision. This article examines domestic and international research and methodologies for developing a three-eye, eight-light source system and fusing head-eye motion data using deep convolution neural dynamics. What followed transpired. Analysis of line of sight characteristics consists of the pupil center, eye image, and pulchin spot center. The irises of humans are identified via face detection. The tracking of the pulchin spot center, face identification, eye image, and pupil center is accomplished through the utilization of a non-wearable three-eye eight light source system, Haar feature, AdaBoost algorithm, and ASM algorithm. To distinguish faces, ASM monitors face feature points in the wide field image and intercepts the human eye image. In order to improve line of sight, this article employs a convolutional neural network to arbitrarily identify faces on-screen. Three random screen orientations result in a classification accuracy of 99% for facial images. The pupil position for the front-facing eye is estimated by a Dual Haar-like feature extractor, followed by the estimation of the pupil center using morphological pixel model and ellipse fitting. Reflection of the purchin Purchin Point is located by a Canny edge detection operator and search algorithm subsequent to threshold processing of the image. In conclusion, the centroid method yields the center coordinates and line of sight of the purchin point. The feature extraction data for head posture presented in this article comprises the following in space axis coordinates: roll, pitch, yaw angle, three-axis position, and speed. Utilize image-based head position evaluation and inertial sensors. Utilizing acceleration interpolation, Picard approximation, and four-order Runge Kutta integral methods, the rotation quaternion is integrated. For tracking purposes, the acceleration interpolation integral converts a head posture quaternion to a rotation angle. Using a face model transformation tailored to the facial feature, the image approach subsequently transforms three-dimensional attributes into two-dimensional space. In order to account for the angle of head posture, the image approach modifies the depth parameters of the model subsequent to converting three-dimensional features to two-dimensional space via facial feature points. Using these two methods, head motion quantification can be enhanced.
- 26 April 2024
- Published by Association for Computing Machinery (ACM)
- p. 59-65
Open AccessResearch and application of gaze direction algorithm based on head movement and eye movement data fusion
Visibility could potentially signify attentiveness. Estimating the line of sight is crucial in numerous fields, including computer vision, human-computer interaction, and psychology. Estimating the line of sight is crucial in numerous fields, including computer vision, human-computer interaction, and psychology. Line of sight estimation is difficult to achieve due to changes in head position and advancements in precision. This article examines domestic and international research and methodologies for developing a three-eye, eight-light source system and fusing head-eye motion data using deep convolution neural dynamics. What followed transpired. Analysis of line of sight characteristics consists of the pupil center, eye image, and pulchin spot center. The irises of humans are identified via face detection. The tracking of the pulchin spot center, face identification, eye image, and pupil center is accomplished through the utilization of a non-wearable three-eye eight light source system, Haar feature, AdaBoost algorithm, and ASM algorithm. To distinguish faces, ASM monitors face feature points in the wide field image and intercepts the human eye image. In order to improve line of sight, this article employs a convolutional neural network to arbitrarily identify faces on-screen. Three random screen orientations result in a classification accuracy of 99% for facial images. The pupil position for the front-facing eye is estimated by a Dual Haar-like feature extractor, followed by the estimation of the pupil center using morphological pixel model and ellipse fitting. Reflection of the purchin Purchin Point is located by a Canny edge detection operator and search algorithm subsequent to threshold processing of the image. In conclusion, the centroid method yields the center coordinates and line of sight of the purchin point. The feature extraction data for head posture presented in this article comprises the following in space axis coordinates: roll, pitch, yaw angle, three-axis position, and speed. Utilize image-based head position evaluation and inertial sensors. Utilizing acceleration interpolation, Picard approximation, and four-order Runge Kutta integral methods, the rotation quaternion is integrated. For tracking purposes, the acceleration interpolation integral converts a head posture quaternion to a rotation angle. Using a face model transformation tailored to the facial feature, the image approach subsequently transforms three-dimensional attributes into two-dimensional space. In order to account for the angle of head posture, the image approach modifies the depth parameters of the model subsequent to converting three-dimensional features to two-dimensional space via facial feature points. Using these two methods, head motion quantification can be enhanced.