Innovation Performance Prediction of University Student Teams Based on Bayesian Networks

Round 1

Reviewer 1 Report

Thank you for submitting your manuscript to Sustainability. It is an interesting read on a relevant topic, but a number of issues require your attention before this paper can be published. 

-- There are many paragraphs to describe the existing methods, e.g., subsections 2.2, 2.4, and 2.5. Brief introductions and citations are fine for them.

-- Also, some basic knowledge of the reliability and validity of a questionnaire  (Row 264-275, Page 7 of 17) is recommended to be left out.

-- The research questions should be clearly presented in the introduction section.

-- The research design should be clearly stated at the beginning of the method section.

-- A independent section of the discussion on the experiment results is recommended.

 

Author Response

Response to Reviewers’ Comments

 

Title: Innovation Performance Prediction of University Student Teams Based on Bayesian Networks

Manuscript Number: Sustainability-2063855

Journal: Sustainability

 

 

Dear reviewers and editors,

 

Thank you so much for your review and advice on our research work.

 

We have revised this paper based on the editor’s and reviewers’ constructive comments and queries thoroughly. The revised parts have been marked with red color in the revised manuscript. A summary of our response to the reviewers’ comments is given below. We hope you find the revisions acceptable.

 

Yours sincerely,

 

Authors

 

 

 

 

 

Reviewer #1:

 

Thank you for submitting your manuscript to Sustainability. It is an interesting read on a relevant topic, but a number of issues require your attention before this paper can be published.

 

1- There are many paragraphs to describe the existing methods, e.g., subsections 2.2, 2.4, and 2.5. Brief introductions and citations are fine for them.

 

Response: Thanks for your comments. Following your suggestion, we have briefly introduced and described the three methods in Subsections 2.2, 2.4 and 2.5. For your convenience, we recall the revised content as follows.

 

2.2 Confirmatory Factor Analysis (Page 4)

In this study, we used Confirmatory Factor Analysis (CFA) to verify the convergent validity and discriminant validity of the scale for research variables, including dualistic learning, task interdependence, knowledge sharing and learning performance. Convergent validity refers to the similarity degree of the same abstract concept or trait measured by different measurement methods. It is usually judged by the standardized factor loadings of each item and the composite reliability. If the value of the standardized factor loading of each item is greater than 0.5 and the value of composite reliability is greater than 0.6, it indicates that the scale has good convergent validity.

Discriminant validity refers to the degree to which a construct is truly different from other constructs according to empirical criteria. In an experiment, if it can be proved statistically that the indicators which are not related to the preset construct are indeed not related to the construct, then the experiment has discriminant validity. In this study, the structural equation model was used to obtain the CFA results of variables. Kline believes that multiple fitting indicators should be used to verify the fitting degree of research samples and theoretical models, and indicators such as, GFI, AGFI, and RMSEA can be used to evaluate the fitting degree of structural equation models [19].

2.4 Confusion Matrix and ROC Curve (Page 5)

To ensure the high accuracy of the prediction model, the accuracy of the BN model must be evaluated.

The Confusion matrix is a basic method to evaluate the reliability of predictive classification models. The column represents the actual situation, while the row represents the prediction result of the classifier. The possible results produced by the classifier are true positive (TP), false positive (FP), false negative (FN), and true negative (TN).

The model accuracy is determined by the following formula (10).

(10)

At the same time, the Bayesian classifier classifies the samples into 0 or 1 according to the classification threshold. Moreover, in these classifiers, different thresholds determine different classification results and classifier evaluation indexes. Considering the above situation, we used the ROC curve and area under the curve (AUC) to measure the model performance. The ROC curve is called the receiver operating characteristic curve, which is commonly used to measure the overall reliability of the classification model. [22]. AUC is the calculated value of the area under the ROC curve, and the value range of AUC is between 0.5 and 1. The greater the AUC value, the better the effect of the machine learning classifier. This method is very convenient and accurate to evaluate the performance of a prediction model.

2.5 Importance Measure (Page 6)

Importance refers to the degree to which the state change of a single or multiple components in the system affects the reliability of the system [27]. It is used to evaluate the relative importance of various factors. Birnbaum first put forward the importance analysis theory in 1969, which is also called probability importance [28].

The classic Birnbaum importance depends on the structure of the system model and has no relationship with the probability of current basic events.

At the moment t, the Birnbaum importance of the i unit is defined as

(11)

Birnbaum importance is obtained from the partial derivative of system reliability to unit reliability . This is the traditional sensitivity analysis method.

For the whole system unit, the importance of Birnbaum can also be expressed as

(12)

where AA represents the probability of the top event when the bottom event i must occur, and represents the probability of the top event when the bottom event i must not occur.

 

2- Also, some basic knowledge of the reliability and validity of a questionnaire (Row 264-275, Page 7 of 17) is recommended to be left out.

 

Response: Thanks for your comments. Following your suggestion, we have removed the redundant content of the reliability and validity of a questionnaire. Such description (highlighted in blue) has been revised on page 6. For your convenience, we recall the revised content as follows.

 

The reliability of the scale is tested by Cronbach’s α value and CR value of the scale. The calculation shows that the Cronbach’s α of all latent variables in this study is greater than 0.7, and the CR value of the composite reliability is greater than 0.6, indicating that the scale has a high level of internal consistency reliability. The scale used in this paper is formed by referring to existing literature, expert interviews, and pre research, which can ensure its content validity. Table 1 shows that the standardized factor loading value of each item is greater than 0.5, and the AVE value of each latent variable is greater than 0.5, indicating that the scale has good convergent validity.

 

3- The research questions should be clearly presented in the introduction section.

 

Response: Thanks for your comments. We have clearly stated the research questions and purposes in introduction Section 1.3. Such description (highlighted in blue) has been revised on page 3. For your convenience, we recall the revised content as follows.

 

Therefore, it is of great theoretical and practical significance to explore the relationship between the main factors affecting university students' team learning and accurately predict the innovation performance of university student teams.

 

4- The research design should be clearly stated at the beginning of the method section.

 

Response: Thanks for your comments. We have clearly stated the research design at the beginning of Section 2. Such description (highlighted in blue) has been revised on page 3. For your convenience, we recall the revised content as follows.

 

This study conducted a questionnaire survey on university students' innovation teams to measure research variables such as dualistic learning, knowledge sharing, and task interdependence, and verified the measurement model and data fitting of this study with confirmatory factor analysis. Based on the sample data, we established two BN prediction models of university student team performance using tree enhanced Bayesian algorithm, and calculated the accuracy of these models. Finally, we quantitatively analyzed the importance degree of each influencing factor and ranked each research variablere according to the experimental results.

 

5- An independent section of the discussion on the experiment results is recommended.

 

Response: Thanks for your comments. We have added an independent section to discuss the experiment results in the paper. Such description (highlighted in blue) has been revised on page 14 and 15. For your convenience, we recall the revised content as follows.

 

5 Discussion (Page 14 and 15)

In this study, we established a prediction model based on the Bayesian network, and designed a method that can analyze the impact of dualistic learning on the innovation performance of university students in different situations. We explored the knowledge construction and learning performance influencing factors of university students' team learning from the perspective of knowledge construction and performance management, and on this basis, further discussed the mechanism of university student teams’ dualistic learning affecting learning performance through empirical research.

The existing literature lacks research on the relationship between team dualistic learning, knowledge sharing, and learning performance from the perspective of predictive modeling. In this study, we established a prediction model based on the Bayesian network, and designed a method that could analyze the impacts of dualistic learning on the innovation performance of university students in different situations. This research focuses on the university student teams, constructs a theoretical model of the relationship between dualistic learning, knowledge sharing, and team learning performance through literature analysis and theoretical analysis, expounds the correlation between variables, and reveals the "black box" of the mechanism of dualistic learning on team learning performance in the context of university students' innovative team learning by using quantitative and qualitative methods, indicating that the improvement of dualistic learning can significantly promote innovation performance.

The theoretical contribution of this study is mainly reflected in two aspects. First, from the perspective of university students' research-based learning, the research results of the enterprise ambidexterity learning theory are expanded to the university students' innovative team learning situation. For university students' learning, the traditional individual level is extended to the team level, the general creativity is shifted to scientific creativity, and the enterprise organizational situation is stretched to the educational organizational situation, which further promotes the integration of team level and scientific dimension and further enriches the theoretical system and research content of organizational learning. Secondly, dividing the ambidexterity learning into exploitative learning and exploratory learning not only explores the direct effect on team task performance directly, but also reveals the indirect effect of ambidexterity learning on team task performance through the mediating effect knowledge sharing behavior, reveals university students' innovation team learning situation of the dual role of team task performance of "black box".

This study has several limitations. Firstly, the data set only contains 228 samples, so more detailed data screening should be carried out for the questionnaires obtained, key information should be identified, and the dataset should be expanded for future research. Secondly, this study doesn’t fully consider the impacts of knowledge sharing and task interdependence on team innovation performance. In the future, on the basis of our calculation of the importance of each influencing factor, we will conduct supplementary research on the mediating role of knowledge sharing and the moderating role of task interdependence.

Author Response File: Author Response.doc

Reviewer 2 Report

The paper is extremely interesting for scientific community and could provide a valuable insight. However, there are some recommendations that can improve it:

I would recommend to restructure the first section in more traditional way (make a division between introduction and literature review (and expand the literature analysis)). 

Methods section should  present information about the sample. Does your sample included students of one university/ nationality? If yes, it should be noted in limitations.

It would be great if you provide some certain recommendations for innovation teams managers, taking in account the results you gained.

 

 

Author Response

Response to Reviewers’ Comments

 

Title: Innovation Performance Prediction of University Student Teams Based on Bayesian Networks

Manuscript Number: Sustainability-2063855

Journal: Sustainability

 

 

Dear reviewers and editors,

 

Thank you so much for your review and advice on our research work.

 

We have revised this paper based on the editor’s and reviewers’ constructive comments and queries thoroughly. The revised parts have been marked with red color in the revised manuscript. A summary of our response to the reviewers’ comments is given below. We hope you find the revisions acceptable.

 

Yours sincerely,

 

Authors

 

 

 

 

 

 

 

 

 

 

 

 

 

Reviewer #2:

 

The paper is extremely interesting for scientific community and could provide a valuable insight. However, there are some recommendations that can improve it:

 

1- I would recommend to restructure the first section in more traditional way (make a division between introduction and literature review (and expand the literature analysis)).

 

Response: Thanks for your comments. Following your suggestion, we have restructured the first section and divided it into three parts, which are background, literature review and motivation. Such division has been revised on page 1,2 and 3. For your convenience, we recall the revised content as follows.

 

1.1. Background

Innovation is the most important driving engine for development and the strategic support for building a modern economic system. As the main force in promoting the development of innovation, university students are increasingly valued by universities and society. Team innovation ability has increasingly become the key factor for organizations to maintain competitive advantage and cope with complex environments. Subsequently, team innovation performance has become a key indicator to measure the team's technological innovation ability and overall performance. Therefore, the research on the prediction of team innovation performance of university students can provide an effective method for the team to improve its ability and gain competitive advantages, so as to stimulate university students’ creativity and enthusiasm.

1.2. Literature Review

Dualistic learning is divided into exploration learning and exploitation learning, in which exploration learning, characterized by "pursuing and acquiring knowledge in new fields, is manifested in the initiative of the organization to create new knowledge, explore new technologies and strategies, and abilities to find new opportunities and new rules." Exploitation learning is characterized by "learning and using existing knowledge" [1], that is, in a mature or new environment, existing knowledge is fully exploited to better apply it to practice, and the original knowledge can be improved and adjusted according to actual needs, so as to improve the implementation efficiency and performance of the established scheme in the specific environment. As for the relationship between dualistic learning and team performance, Katila and Ahuja [2] found that the combination of the two learning methods can enhance the viability of the organization, improve the financial performance of the enterprise, and enhance the ability of organizational learning and innovation. Colbert [3] believes that the competitive advantage generated by the interaction of two learning methods is better than the competitive advantage obtained by carrying out one learning method alone. Some scholars have also analyzed the characteristics of complex network structures and believed that a dualistic balanced network structure can better promote organizational performance [4]. However, the impacts of dualistic capability on organizational performance is flexible [5], and the change in network situation may significantly affect organizational performance. Dualistic learning plays a more important role in improving the performance of enterprise alliances with higher network centrality. It can also help enterprise alliances with fewer structural holes to achieve better performance. Although the relationship between dualistic learning and team performance has been widely studied, it is mainly concentrated in the industrial and business sectors. At present, scholars have empirically studied the impacts of dualistic learning, knowledge sharing, member heterogeneity, and their influencing factors on team performance in an enterprise [6–8]. The results from these researches have important guiding significance for human resource management practice and innovation management practice in the enterprise field. However, there are substantial differences between university student teams and enterprise teams. There are also substantial differences in learning objects and innovation pursuits. This study focuses on the theoretical issue of "the mechanism of dualistic learning impacts on university student team learning performance" to explore the impacts of dualistic learning on innovation performance in university student teams. Hopefully, this research will enrich the existing theoretical connotation to a certain extent and promote the development of relevant theoretical research.

With the rapid development of computers and chips, the limitation of computing power is constantly broken through, and a large number of Artificial Intelligence (AI) methods, including Bayesian methods, Support Vector Machine (SVM), integrated learning, and deep learning, are emerging and gradually applied to data mining and predictive modeling [9-13]. Compared with the traditional generalized linear model, the Bayesian network has fewer restrictions on modeling variables and is more robust to the nonlinear relationship between variables. It can better fit the data distribution and give more accurate prediction results. At present, Bayesian network (BN) is widely used in industrial applications, financial analysis, and military applications. Sun et al. [14] proposed an evolutionary algorithm based on BN to optimize the task allocation strategy, which to some extent solved the response delay when the industrial Internet of Things is used to monitor industrial plants. Bashar et al. [15] proposed three new network management solutions based on BN, which to some extent meet the requirements of automation and efficient management of telecommunications networks. Mahadevan et al. [16] put forward a structural reliability evaluation method combining BN, and also considered the multiple failure sequence characteristics of large structures and the correlation between element limit states. The effectiveness of this evaluation method was verified by comparison with the traditional series parallel reliability method. In order to control the quality risk of cold storage construction projects, Song et al. [17] constructed a BN based cold storage construction quality risk assessment model, and obtained key quality risk factors through reverse reasoning analysis and sensitivity analysis. Johnson et al. [18] created a Math App to provide students in Emirates in grade 6 with the opportunity to explore mathematics, and then used BN to examine the educational implications and explore the impacts of different factors on their enthusiasm for learning mathematics.

 

2- Methods section should present information about the sample. Does your sample included students of one university/ nationality? If yes, it should be noted in limitations.

 

Response: Thanks for your comments. Actually, in the research samples, the student teams come from three types of universities or institutions: world-class universities, world-class disciplines and other universities. Following your suggestion, we have added information on the source and proportion of 228 samples in Section 3.2. Such description (highlighted in blue) has been revised on page 8. For your convenience, we recall the revised content as follows.

 

Among the respondents, 121 teams (53.1%) come from world-class universities, 24 teams (10.5%) come from world-class disciplines, and 83 teams (36.4%) come from other universities.

 

3- It would be great if you provide some certain recommendations for innovation teams managers, taking in account the results you gained.

 

Response: Thanks for your comments. Based on the experimental results, we have added some management strategies and practical enlightenment for innovation teams managers on page 15 and 16 in the revised paper (highlighted in blue). For your convenience, we recall the revised content as follows.

The practical enlightenment of this study are as follow. 1) From the perspective of policy making by innovation teams managers, the innovation and entrepreneurship management departments of colleges and universities should actively build an academic community for university students' team knowledge sharing, formulate a scientific and reasonable system, stimulate students' knowledge sharing motivation, improve students' knowledge sharing ability, promote university students' knowledge resource sharing, and then improve the team's innovation output. 2) From the perspective of university students’ team building, we should pay attention to establishing good interpersonal interaction and mutual learning relationship with other members, constantly optimise the team knowledge sharing environment, promote the knowledge circulation flow within the team, further create a strong academic atmosphere to stimulate college students' learning motivation and enthusiasm. 3) From the perspective of individual students, it is necessary to associate expectation of success with teams’ subjective task value and pay more attention to the selection and application of learning strategies, so as to promote their teams to achieve higher task performance.

Author Response File: Author Response.doc

Reviewer 3 Report

 

This manuscript is interesting, addresses an important issue. One gets the impression that the authors pay a lot of attention to statistics (which is welcome), but there is a lack of theoretical justification and description of the research design. This prompted the following comments..

1)     Structure of the paper. I’d strongly recommend separating the introduction and the theoretical part. I would strongly suggest rewriting the introduction by summarizing the existing knowledge on the topic and revealing the research gap, disclosing the novelty element. Identify the contribution of your study and explain why that contribution is important for advancing the field. Next, in the literature review (theoretical part), define the main constructs, explain the relationships between them. Make sure it relates to the survey questionnaire.

2)     Authors state, that “<..>mature scale developed by domestic and foreign scholars to  ensure the validity and reliability of the questionnaire “ [line 113-114]. Which measures, scales (or separate questions, their groups) were used? What authors? Please provide in detail.

3)     What was the study sample, how was it designed? Now  it is a small remark in line 428 (in conclusions)

4)     I missed the discussion on the research results. To what extent they are inline to the results of other scholars' research? The authors claim that "Many studies have been conducted on the impact of dualistic learning, knowledge sharing, member heterogeneity, and their influencing factors about team performance in enterprises" [lines 9-10]. What did your research reveal? The authors are advised to develop a clear link between literature and empirical study to show how this study fits in within the existing literature. Later, identify the contribution of your study and explain why that contribution is important for advancing the field.

Author Response

Response to Reviewers’ Comments

 

Title: Innovation Performance Prediction of University Student Teams Based on Bayesian Networks

Manuscript Number: Sustainability-2063855

Journal: Sustainability

 

 

Dear reviewers and editors,

 

Thank you so much for your review and advice on our research work.

 

We have revised this paper based on the editor’s and reviewers’ constructive comments and queries thoroughly. The revised parts have been marked with red color in the revised manuscript. A summary of our response to the reviewers’ comments is given below. We hope you find the revisions acceptable.

 

Yours sincerely,

 

Authors

 

 

 

 

 

Reviewer #3:

 

This manuscript is interesting, addresses an important issue. One gets the impression that the authors pay a lot of attention to statistics (which is welcome), but there is a lack of theoretical justification and description of the research design. This prompted the following comments.

 

• Structure of the paper. I’d strongly recommend separating the introduction and the theoretical part. I would strongly suggest rewriting the introduction by summarizing the existing knowledge on the topic and revealing the research gap, disclosing the novelty element. Identify the contribution of your study and explain why that contribution is important for advancing the field. Next, in the literature review (theoretical part), define the main constructs, explain the relationships between them. Make sure it relates to the survey questionnaire.

 

Response: Thanks for your comments. Following your suggestion, we have restructured the introduction section and made necessary revisions to the section. We have divided it into three parts. In subsection 1.1(the first part), we have introduced the research background and implied the significance and novelty of using Bayesian network model to predict performance in this research. In subsection 1.2(the second part), we have reviewed the literature and summarized the existing knowledge on dualistic learning, innovation performance and bayesian network. In subsection 1.3, we have clarified our research motivation, revealed the research gap, and explained the practical significance of using Bayesian network model to predict student teams’ innovation performance.Such description (highlighted in blue) has been revised on page 1,2 and 3. For your convenience, we recall the revised content as follows.

 

1.1. Background

Innovation is the most important driving engine for development and the strategic support for building a modern economic system. As the main force in promoting the development of innovation, university students are increasingly valued by universities and society. Team innovation ability has increasingly become the key factor for organizations to maintain competitive advantage and cope with complex environments. Subsequently, team innovation performance has become a key indicator to measure the team's technological innovation ability and overall performance. Therefore, the research on the prediction of team innovation performance of university students can provide an effective method for the team to improve its ability and gain competitive advantages, so as to stimulate university students’ creativity and enthusiasm.

1.2. Literature Review

Dualistic learning is divided into exploration learning and exploitation learning, in which exploration learning, characterized by "pursuing and acquiring knowledge in new fields, is manifested in the initiative of the organization to create new knowledge, explore new technologies and strategies, and abilities to find new opportunities and new rules." Exploitation learning is characterized by "learning and using existing knowledge" [1], that is, in a mature or new environment, existing knowledge is fully exploited to better apply it to practice, and the original knowledge can be improved and adjusted according to actual needs, so as to improve the implementation efficiency and performance of the established scheme in the specific environment. As for the relationship between dualistic learning and team performance, Katila and Ahuja [2] found that the combination of the two learning methods can enhance the viability of the organization, improve the financial performance of the enterprise, and enhance the ability of organizational learning and innovation. Colbert [3] believes that the competitive advantage generated by the interaction of two learning methods is better than the competitive advantage obtained by carrying out one learning method alone. Some scholars have also analyzed the characteristics of complex network structures and believed that a dualistic balanced network structure can better promote organizational performance [4]. However, the impacts of dualistic capability on organizational performance is flexible [5], and the change in network situation may significantly affect organizational performance. Dualistic learning plays a more important role in improving the performance of enterprise alliances with higher network centrality. It can also help enterprise alliances with fewer structural holes to achieve better performance. Although the relationship between dualistic learning and team performance has been widely studied, it is mainly concentrated in the industrial and business sectors. At present, scholars have empirically studied the impacts of dualistic learning, knowledge sharing, member heterogeneity, and their influencing factors on team performance in an enterprise [6–8]. The results from these researches have important guiding significance for human resource management practice and innovation management practice in the enterprise field. However, there are substantial differences between university student teams and enterprise teams. There are also substantial differences in learning objects and innovation pursuits. This study focuses on the theoretical issue of "the mechanism of dualistic learning impacts on university student team learning performance" to explore the impacts of dualistic learning on innovation performance in university student teams. Hopefully, this research will enrich the existing theoretical connotation to a certain extent and promote the development of relevant theoretical research.

With the rapid development of computers and chips, the limitation of computing power is constantly broken through, and a large number of Artificial Intelligence (AI) methods, including Bayesian methods, Support Vector Machine (SVM), integrated learning, and deep learning, are emerging and gradually applied to data mining and predictive modeling [9–13]. Compared with the traditional generalized linear model, the Bayesian network has fewer restrictions on modeling variables and is more robust to the nonlinear relationship between variables. It can better fit the data distribution and give more accurate prediction results. At present, Bayesian network (BN) is widely used in industrial applications, financial analysis, and military applications. Sun et al. [14] proposed an evolutionary algorithm based on BN to optimize the task allocation strategy, which to some extent solved the response delay when the industrial Internet of Things is used to monitor industrial plants. Bashar et al. [15] proposed three new network management solutions based on BN, which to some extent meet the requirements of automation and efficient management of telecommunications networks. Mahadevan et al. [16] put forward a structural reliability evaluation method combining BN, and also considered the multiple failure sequence characteristics of large structures and the correlation between element limit states. The effectiveness of this evaluation method was verified by comparison with the traditional series parallel reliability method. In order to control the quality risk of cold storage construction projects, Song et al. [17] constructed a BN based cold storage construction quality risk assessment model, and obtained key quality risk factors through reverse reasoning analysis and sensitivity analysis. Johnson et al. [18] created a Math App to provide students in Emirates in grade 6 with the opportunity to explore mathematics, and then used BN to examine the educational implications and explore the impacts of different factors on their enthusiasm for learning mathematics.

1.3. Motiation

To sum up, the existing literature lacks research on the relationship between dualistic learning, knowledge sharing, and learning performance of university student teams from the perspective of organizational learning and knowledge management. Meanwhile, research on innovation performance using predictive models is relatively scarce. Therefore, it is of great theoretical and practical significance to explore the relationship between the main factors affecting university students' team learning and accurately predict the innovation performance of university student teams.

In this study, we aim to use BN to establish two prediction models, which compare the impacts of different levels of dualistic learning on university student teams’ learning performance, and explore how the mechanism of university student teams’ dualistic learning affect their learning performance.

This paper starts by introducing the technical methods and evaluation parameters used in the research (Section 2). And the prediction models of two groups of different classification targets are established based on tree enhanced Bayesian network (Section 3). In section 4, we explore the impacts of dualistic learning on innovation performance. In Section 5, the research perspective and modeling methods are presented and discussed. Finally, it concludes with limitations and recommendations to future researchers and practitioners.

 

• Authors state, that “<..>mature scale developed by domestic and foreign scholars to ensure the validity and reliability of the questionnaire “ [line 113-114]. Which measures, scales (or separate questions, their groups) were used? What authors? Please provide in detail.

 

Response: Thanks for your comments. Following your suggestion, we have made a detailed statement and explanation of the scales used in the research and the authors of these scales in Section 2.1. Such description (highlighted in blue) has been revised on page 3. For your convenience, we recall the revised content as follows.

 

Among them, the ambidexterity learning scale [19] includes two dimensions of exploitative learning and exploratory learning, both of which are developed by Atuahene-Gema and Murray. Each dimensions includes three measurement items; The knowledge sharing behavior scale [20] is developed by Collins and translated by Tian Lifa, including four dimensions. The team task performance scale adopts the scale [21] developed by Chen Wei and others, which has four dimensions. The task interdependence scale [22] is developed by Liden, Wayne and Broadway, and translated by Liu Jun, with a total of three items.

 

• What was the study sample, how was it designed? Now it is a small remark in line 428 (in conclusions)

 

Response: Thanks for your comments. Following your suggestion, we have added detailed information on the source and proportion of 228 samples in Conclusion (Section 6). In addition, we have clearly stated our research design in this paragraph. Such description (highlighted in blue) has been revised on page 15. For your convenience, we recall the revised content as follows.

 

We obtained 228 samples of university student teams from three types of universities. Among them, 121 samples (53.1%) were obtained from world-class universities, 24 samples (10.5%) were obtained from world-class disciplines, and 83 samples (36.4%) were obtained from other universities. We evaluated these samples and established two BN prediction models of university student team performance using tree enhanced Bayesian algorithm.

 

• I missed the discussion on the research results. To what extent they are inline to the results of other scholars' research? The authors claim that "Many studies have been conducted on the impact of dualistic learning, knowledge sharing, member heterogeneity, and their influencing factors about team performance in enterprises" [lines 9-10]. What did your research reveal? The authors are advised to develop a clear link between literature and empirical study to show how this study fits in within the existing literature. Later, identify the contribution of your study and explain why that contribution is important for advancing the field.

 

Response: Thanks for your comments. First, we have added an independent section to discuss the experimental results. In this section, we have compared the research with other similar research, indicating that although the methods are quite different, these studies have reached some similar conclusions. Such description (highlighted in blue) has been revised on page 14 (the second paragraph of Section 5). For your convenience, we recall the revised content as follows.

 

It is basically consistent with the results of other scholars’ research, that is, dualistic learning, as well as knowledge sharing, can positively affect innovation performance or task performance. Although the research objects and methods are different, there are many similarities of conclusions between these studies.

 

Second, through literature analysis, we found the shortage of existing literature, and elaborated our research motivation and purpose: To reveal the mechanism of dualistic learning on team learning performance in the context of university students' innovative team learning and to predict the innovation performance of university students' teams by empirical study. Such description (highlighted in blue) has been revised on page 14 (the second paragraph of Section 5). For your convenience, we recall the revised content as follows.

 

The existing literature lacks research on the relationship between team dualistic learning, knowledge sharing, and learning performance from the perspective of predictive modeling. Therefore, we aimed to use statistical methods and importance measure theory to supplement the shortage of existing literature. This research focuses on the university student teams, constructs a theoretical model of the relationship between dualistic learning, knowledge sharing, and team learning performance and reveals the "black box" of the mechanism of dualistic learning on team learning performance in the context of university students' innovative team learning by using quantitative and qualitative methods, indicating that the improvement of dualistic learning can significantly promote innovation performance.

 

Finally, following your suggestion, we have added 2 main theoretical contributions in the second paragraph of Section 5. For the one hand, the research results of enterprise dualistic learning theory are extended to the innovative team learning situation of college students. For the other hand, the study reveals the mechanism of dualistic learning affects university student teams’ learning performance. Such description (highlighted in blue) has been revised in the third paragraph of Section 5. For your convenience, we recall the revised content as follows.

 

The theoretical contribution of this study is mainly reflected in two aspects. First, from the perspective of university students' research-based learning, the research results of the enterprise ambidexterity learning theory are expanded to the university students' innovative team learning situation. For university students' learning, the traditional individual level is extended to the team level, the general creativity is shifted to scientific creativity, and the enterprise organizational situation is stretched to the educational organizational situation, which further promotes the integration of team level and scientific dimension and further enriches the theoretical system and research content of organizational learning. Secondly, dividing the ambidexterity learning into exploitative learning and exploratory learning not only explores the direct effect on team task performance directly, but also reveals the indirect effect of ambidexterity learning on team task performance through the mediating effect knowledge sharing behavior, reveals university students' innovation team learning situation of the dual role of team task performance of "black box".

 

Author Response File: Author Response.doc

Round 2

Reviewer 1 Report

The authors have addressed my concerns. 

English language and style may require minor spell checks. 

Author Response

Please see the attachment.

Author Response File: Author Response.doc

Reviewer 3 Report

 

Thank you for the efforts in improving the paper. However, I would like to draw attention to the comments expressed earlier and new comments raised by authors' corrections. 

 

Structure of the paper. it is not enough to simply place the title between paragraphs.

 

1)     The introduction must contain certain information. The introduction focuses on the practical issues that are revealed. I would suggest extending the introduction by summarizing the existing knowledge on the topic and providing a research question. Now is not clear what is the scientific issue of the paper. An introduction should inform about the issue under investigation: 1) What is the topic of research question, and why is it important to theory and practice, 2) Summarize existing knowledge on the topic and provide the gap. 3) what is theoretical contribution of this study? Approximately 1 page

2)Theoretical background. „1.3. Motiation“ (Line 98)  What do have in mind  by “Motiation” ?. Any way, such a text is a part of an introduction, but not part of a Literature review. The theoretical part must end with specific theoretical solutions. Please define the main constructs, explain the relationships between them. Make sure it relates to the survey questionnaire.

I would highly recommend that suggestions be revised by making essencial content rather than cosmetic/ technical corrections

Author Response

Please see the attachment.

Author Response File: Author Response.doc

Round 3

Reviewer 3 Report

Thank you for the efforts in improving the paper.

Author Response

Please see the attachment.

Author Response File: Author Response.doc