In geotechnical and tunneling engineering,accurately determining the mechanical properties of jointed rock holds great significance for project safety assessments.Peak shear strength(PSS),being the paramount mechanica...In geotechnical and tunneling engineering,accurately determining the mechanical properties of jointed rock holds great significance for project safety assessments.Peak shear strength(PSS),being the paramount mechanical property of joints,has been a focal point in the research field.There are limitations in the current peak shear strength(PSS)prediction models for jointed rock:(i)the models do not comprehensively consider various influencing factors,and a PSS prediction model covering seven factors has not been established,including the sampling interval of the joints,the surface roughness of the joints,the normal stress,the basic friction angle,the uniaxial tensile strength,the uniaxial compressive strength,and the joint size for coupled joints;(ii)the datasets used to train the models are relatively limited;and(iii)there is a controversy regarding whether compressive or tensile strength should be used as the strength term among the influencing factors.To overcome these limitations,we developed four machine learning models covering these seven influencing factors,three relying on Support Vector Regression(SVR)with different kernel functions(linear,polynomial,and Radial Basis Function(RBF))and one using deep learning(DL).Based on these seven influencing factors,we compiled a dataset comprising the outcomes of 493 published direct shear tests for the training and validation of these four models.We compared the prediction performance of these four machine learning models with Tang’s and Tatone’s models.The prediction errors of Tang’s and Tatone’s models are 21.8%and 17.7%,respectively,while SVR_linear is at 16.6%,SVR_poly is at 14.0%,and SVR_RBF is at 12.1%.DL outperforms the two existing models with only an 8.5%error.Additionally,we performed shear tests on granite joints to validate the predictive capability of the DL-based model.With the DL approach,the results suggest that uniaxial tensile strength is recommended as the material strength term in the PSS model for more reliable outcomes.展开更多
Emotion cause extraction(ECE)task that aims at extracting potential trigger events of certain emotions has attracted extensive attention recently.However,current work neglects the implicit emotion expressed without an...Emotion cause extraction(ECE)task that aims at extracting potential trigger events of certain emotions has attracted extensive attention recently.However,current work neglects the implicit emotion expressed without any explicit emotional keywords,which appears more frequently in application scenarios.The lack of explicit emotion information makes it extremely hard to extract emotion causes only with the local context.Moreover,an entire event is usually across multiple clauses,while existing work merely extracts cause events at clause level and cannot effectively capture complete cause event information.To address these issues,the events are first redefined at the tuple level and a span-based tuple-level algorithm is proposed to extract events from different clauses.Based on it,a corpus for implicit emotion cause extraction that tries to extract causes of implicit emotions is constructed.The authors propose a knowledge-enriched jointlearning model of implicit emotion recognition and implicit emotion cause extraction tasks(KJ-IECE),which leverages commonsense knowledge from ConceptNet and NRC_VAD to better capture connections between emotion and corresponding cause events.Experiments on both implicit and explicit emotion cause extraction datasets demonstrate the effectiveness of the proposed model.展开更多
Multimodal Sentiment Analysis(SA)is gaining popularity due to its broad application potential.The existing studies have focused on the SA of single modalities,such as texts or photos,posing challenges in effectively h...Multimodal Sentiment Analysis(SA)is gaining popularity due to its broad application potential.The existing studies have focused on the SA of single modalities,such as texts or photos,posing challenges in effectively handling social media data with multiple modalities.Moreover,most multimodal research has concentrated on merely combining the two modalities rather than exploring their complex correlations,leading to unsatisfactory sentiment classification results.Motivated by this,we propose a new visualtextual sentiment classification model named Multi-Model Fusion(MMF),which uses a mixed fusion framework for SA to effectively capture the essential information and the intrinsic relationship between the visual and textual content.The proposed model comprises three deep neural networks.Two different neural networks are proposed to extract the most emotionally relevant aspects of image and text data.Thus,more discriminative features are gathered for accurate sentiment classification.Then,a multichannel joint fusion modelwith a self-attention technique is proposed to exploit the intrinsic correlation between visual and textual characteristics and obtain emotionally rich information for joint sentiment classification.Finally,the results of the three classifiers are integrated using a decision fusion scheme to improve the robustness and generalizability of the proposed model.An interpretable visual-textual sentiment classification model is further developed using the Local Interpretable Model-agnostic Explanation model(LIME)to ensure the model’s explainability and resilience.The proposed MMF model has been tested on four real-world sentiment datasets,achieving(99.78%)accuracy on Binary_Getty(BG),(99.12%)on Binary_iStock(BIS),(95.70%)on Twitter,and(79.06%)on the Multi-View Sentiment Analysis(MVSA)dataset.These results demonstrate the superior performance of our MMF model compared to single-model approaches and current state-of-the-art techniques based on model evaluation criteria.展开更多
To solve the problem of missing many valid triples in knowledge graphs(KGs),a novel model based on a convolutional neural network(CNN)called ConvKG is proposed,which employs a joint learning strategy for knowledge gra...To solve the problem of missing many valid triples in knowledge graphs(KGs),a novel model based on a convolutional neural network(CNN)called ConvKG is proposed,which employs a joint learning strategy for knowledge graph completion(KGC).Related research work has shown the superiority of convolutional neural networks(CNNs)in extracting semantic features of triple embeddings.However,these researches use only one single-shaped filter and fail to extract semantic features of different granularity.To solve this problem,ConvKG exploits multi-shaped filters to co-convolute on the triple embeddings,joint learning semantic features of different granularity.Different shaped filters cover different sizes on the triple embeddings and capture pairwise interactions of different granularity among triple elements.Experimental results confirm the strength of joint learning,and compared with state-of-the-art CNN-based KGC models,ConvKG achieves the better mean rank(MR)and Hits@10 metrics on dataset WN18 RR,and the better MR on dataset FB15k-237.展开更多
A distributed reinforcement learning(RL)based resource management framework is proposed for a mobile edge computing(MEC)system with both latency-sensitive and latency-insensitive services.We investigate joint optimiza...A distributed reinforcement learning(RL)based resource management framework is proposed for a mobile edge computing(MEC)system with both latency-sensitive and latency-insensitive services.We investigate joint optimization of both computing and radio resources to achieve efficient on-demand matches of multi-dimensional resources and diverse requirements of users.A multi-objective integer programming problem is formulated by two subproblems,i.e.,access point(AP)selection and subcarrier allocation,which can be solved jointly by our proposed distributed RL-based approach with a heuristic iteration algorithm.The proposed algorithm allows for the reduction in complexity since each user needs to consider only its own selection of AP without knowing full global information.Simulation results show that our algorithm can achieve near-optimal performance while reducing computational complexity significantly.Compared with other algorithms that only optimize either of the two sub-problems,the proposed algorithm can serve more users with much less power consumption and content delivery latency.展开更多
In the era of Big data,learning discriminant feature representation from network traffic is identified has as an invariably essential task for improving the detection ability of an intrusion detection system(IDS).Owin...In the era of Big data,learning discriminant feature representation from network traffic is identified has as an invariably essential task for improving the detection ability of an intrusion detection system(IDS).Owing to the lack of accurately labeled network traffic data,many unsupervised feature representation learning models have been proposed with state-of-theart performance.Yet,these models fail to consider the classification error while learning the feature representation.Intuitively,the learnt feature representation may degrade the performance of the classification task.For the first time in the field of intrusion detection,this paper proposes an unsupervised IDS model leveraging the benefits of deep autoencoder(DAE)for learning the robust feature representation and one-class support vector machine(OCSVM)for finding the more compact decision hyperplane for intrusion detection.Specially,the proposed model defines a new unified objective function to minimize the reconstruction and classification error simultaneously.This unique contribution not only enables the model to support joint learning for feature representation and classifier training but also guides to learn the robust feature representation which can improve the discrimination ability of the classifier for intrusion detection.Three set of evaluation experiments are conducted to demonstrate the potential of the proposed model.First,the ablation evaluation on benchmark dataset,NSL-KDD validates the design decision of the proposed model.Next,the performance evaluation on recent intrusion dataset,UNSW-NB15 signifies the stable performance of the proposed model.Finally,the comparative evaluation verifies the efficacy of the proposed model against recently published state-of-the-art methods.展开更多
Artificial intelligent based dialog systems are getting attention from both business and academic communities.The key parts for such intelligent chatbot systems are domain classification,intent detection,and named ent...Artificial intelligent based dialog systems are getting attention from both business and academic communities.The key parts for such intelligent chatbot systems are domain classification,intent detection,and named entity recognition.Various supervised,unsupervised,and hybrid approaches are used to detect each field.Such intelligent systems,also called natural language understanding systems analyze user requests in sequential order:domain classification,intent,and entity recognition based on the semantic rules of the classified domain.This sequential approach propagates the downstream error;i.e.,if the domain classification model fails to classify the domain,intent and entity recognition fail.Furthermore,training such intelligent system necessitates a large number of user-annotated datasets for each domain.This study proposes a single joint predictive deep neural network framework based on long short-term memory using only a small user-annotated dataset to address these issues.It investigates value added by incorporating unlabeled data from user chatting logs into multi-domain spoken language understanding systems.Systematic experimental analysis of the proposed joint frameworks,along with the semi-supervised multi-domain model,using open-source annotated and unannotated utterances shows robust improvement in the predictive performance of the proposed multi-domain intelligent chatbot over a base joint model and joint model based on adversarial learning.展开更多
Due to the many types of distributed denial-of-service attacks(DDoS)attacks and the large amount of data generated,it becomes a chal-lenge to manage and apply the malicious behavior knowledge generated by DDoS attacks...Due to the many types of distributed denial-of-service attacks(DDoS)attacks and the large amount of data generated,it becomes a chal-lenge to manage and apply the malicious behavior knowledge generated by DDoS attacks.We propose a malicious behavior knowledge base framework for DDoS attacks,which completes the construction and application of a multi-domain malicious behavior knowledge base.First,we collected mali-cious behavior traffic generated by five mainstream DDoS attacks.At the same time,we completed the knowledge collection mechanism through data pre-processing and dataset design.Then,we designed a malicious behavior category graph and malicious behavior structure graph for the characteristic information and spatial structure of DDoS attacks and completed the knowl-edge learning mechanism using a graph neural network model.To protect the data privacy of multiple multi-domain malicious behavior knowledge bases,we implement the knowledge-sharing mechanism based on federated learning.Finally,we store the constructed knowledge graphs,graph neural network model,and Federated model into the malicious behavior knowledge base to complete the knowledge management mechanism.The experimental results show that our proposed system architecture can effectively construct and apply the malicious behavior knowledge base,and the detection capability of multiple DDoS attacks occurring in the network reaches above 0.95,while there exists a certain anti-interference capability for data poisoning cases.展开更多
中文电子病历实体关系抽取是构建医疗知识图谱,服务下游子任务的重要基础。目前,中文电子病例进行实体关系抽取仍存在因医疗文本关系复杂、实体密度大而造成医疗名词识别不准确的问题。针对这一问题,提出了基于对抗学习与多特征融合的...中文电子病历实体关系抽取是构建医疗知识图谱,服务下游子任务的重要基础。目前,中文电子病例进行实体关系抽取仍存在因医疗文本关系复杂、实体密度大而造成医疗名词识别不准确的问题。针对这一问题,提出了基于对抗学习与多特征融合的中文电子病历实体关系联合抽取模型AMFRel(adversarial learning and multi-feature fusion for relation triple extraction),提取电子病历的文本和词性特征,得到融合词性信息的编码向量;利用编码向量联合对抗训练产生的扰动生成对抗样本,抽取句子主语;利用信息融合模块丰富文本结构特征,并根据特定的关系信息抽取出相应的宾语,得到医疗文本的三元组。采用CHIP2020关系抽取数据集和糖尿病数据集进行实验验证,结果显示:AMFRel在CHIP2020关系抽取数据集上的Precision为63.922%,Recall为57.279%,F1值为60.418%;在糖尿病数据集上的Precision、Recall和F1值分别为83.914%,67.021%和74.522%,证明了该模型的三元组抽取性能优于其他基线模型。展开更多
目前,大部分将知识图谱引入推荐系统的方法只是将已知的表层知识图谱实体进行引入,没有对图谱的内在关系进行预测和挖掘,因此无法利用知识图谱中的隐藏关系。针对上述问题,提出联合学习推荐模型E-TUP(enhance towards understanding of ...目前,大部分将知识图谱引入推荐系统的方法只是将已知的表层知识图谱实体进行引入,没有对图谱的内在关系进行预测和挖掘,因此无法利用知识图谱中的隐藏关系。针对上述问题,提出联合学习推荐模型E-TUP(enhance towards understanding of user preference),使用E-CP(enhance canonical polyadic)进行知识图谱补全并将完整信息进行传递。利用储存空间负采样方法,将优质负例三元组进行存储,并随训练过程进行更新,以提高知识图谱补全中负例三元组的质量。链接预测实验结果显示,储存空间方法使E-TUP模型链接预测准确率对比现有模型最高提升10.3%。在MovieLens-1m和DBbook2014数据集上进行推荐实验,在多个评价指标上取得最佳结果,对比现有模型实现最高5.5%的提升,表明E-TUP可以有效利用知识图谱中的隐藏关系提高模型推荐准确率。基于汽车维修数据进行推荐实验,结果表明E-TUP可以有效推荐相关知识。展开更多
基金supported by the National Key Research and Development Program of China(2022YFC3080100)the National Natural Science Foundation of China(Nos.52104090,52208328 and 12272353)+1 种基金the Open Fund of Anhui Province Key Laboratory of Building Structure and Underground Engineering,Anhui Jianzhu University(No.KLBSUE-2022-06)the Open Research Fund of Key Laboratory of Construction and Safety of Water Engineering of the Ministry of Water Resources,China Institute of Water Resources and Hydropower Research(Grant No.IWHR-ENGI-202302)。
文摘In geotechnical and tunneling engineering,accurately determining the mechanical properties of jointed rock holds great significance for project safety assessments.Peak shear strength(PSS),being the paramount mechanical property of joints,has been a focal point in the research field.There are limitations in the current peak shear strength(PSS)prediction models for jointed rock:(i)the models do not comprehensively consider various influencing factors,and a PSS prediction model covering seven factors has not been established,including the sampling interval of the joints,the surface roughness of the joints,the normal stress,the basic friction angle,the uniaxial tensile strength,the uniaxial compressive strength,and the joint size for coupled joints;(ii)the datasets used to train the models are relatively limited;and(iii)there is a controversy regarding whether compressive or tensile strength should be used as the strength term among the influencing factors.To overcome these limitations,we developed four machine learning models covering these seven influencing factors,three relying on Support Vector Regression(SVR)with different kernel functions(linear,polynomial,and Radial Basis Function(RBF))and one using deep learning(DL).Based on these seven influencing factors,we compiled a dataset comprising the outcomes of 493 published direct shear tests for the training and validation of these four models.We compared the prediction performance of these four machine learning models with Tang’s and Tatone’s models.The prediction errors of Tang’s and Tatone’s models are 21.8%and 17.7%,respectively,while SVR_linear is at 16.6%,SVR_poly is at 14.0%,and SVR_RBF is at 12.1%.DL outperforms the two existing models with only an 8.5%error.Additionally,we performed shear tests on granite joints to validate the predictive capability of the DL-based model.With the DL approach,the results suggest that uniaxial tensile strength is recommended as the material strength term in the PSS model for more reliable outcomes.
基金National Natural Science Foundation of China,Grant/Award Numbers:61671064,61732005National Key Research&Development Program,Grant/Award Number:2018YFC0831700。
文摘Emotion cause extraction(ECE)task that aims at extracting potential trigger events of certain emotions has attracted extensive attention recently.However,current work neglects the implicit emotion expressed without any explicit emotional keywords,which appears more frequently in application scenarios.The lack of explicit emotion information makes it extremely hard to extract emotion causes only with the local context.Moreover,an entire event is usually across multiple clauses,while existing work merely extracts cause events at clause level and cannot effectively capture complete cause event information.To address these issues,the events are first redefined at the tuple level and a span-based tuple-level algorithm is proposed to extract events from different clauses.Based on it,a corpus for implicit emotion cause extraction that tries to extract causes of implicit emotions is constructed.The authors propose a knowledge-enriched jointlearning model of implicit emotion recognition and implicit emotion cause extraction tasks(KJ-IECE),which leverages commonsense knowledge from ConceptNet and NRC_VAD to better capture connections between emotion and corresponding cause events.Experiments on both implicit and explicit emotion cause extraction datasets demonstrate the effectiveness of the proposed model.
文摘Multimodal Sentiment Analysis(SA)is gaining popularity due to its broad application potential.The existing studies have focused on the SA of single modalities,such as texts or photos,posing challenges in effectively handling social media data with multiple modalities.Moreover,most multimodal research has concentrated on merely combining the two modalities rather than exploring their complex correlations,leading to unsatisfactory sentiment classification results.Motivated by this,we propose a new visualtextual sentiment classification model named Multi-Model Fusion(MMF),which uses a mixed fusion framework for SA to effectively capture the essential information and the intrinsic relationship between the visual and textual content.The proposed model comprises three deep neural networks.Two different neural networks are proposed to extract the most emotionally relevant aspects of image and text data.Thus,more discriminative features are gathered for accurate sentiment classification.Then,a multichannel joint fusion modelwith a self-attention technique is proposed to exploit the intrinsic correlation between visual and textual characteristics and obtain emotionally rich information for joint sentiment classification.Finally,the results of the three classifiers are integrated using a decision fusion scheme to improve the robustness and generalizability of the proposed model.An interpretable visual-textual sentiment classification model is further developed using the Local Interpretable Model-agnostic Explanation model(LIME)to ensure the model’s explainability and resilience.The proposed MMF model has been tested on four real-world sentiment datasets,achieving(99.78%)accuracy on Binary_Getty(BG),(99.12%)on Binary_iStock(BIS),(95.70%)on Twitter,and(79.06%)on the Multi-View Sentiment Analysis(MVSA)dataset.These results demonstrate the superior performance of our MMF model compared to single-model approaches and current state-of-the-art techniques based on model evaluation criteria.
基金Supported by the National Natural Science Foundation of China(No.61876144)。
文摘To solve the problem of missing many valid triples in knowledge graphs(KGs),a novel model based on a convolutional neural network(CNN)called ConvKG is proposed,which employs a joint learning strategy for knowledge graph completion(KGC).Related research work has shown the superiority of convolutional neural networks(CNNs)in extracting semantic features of triple embeddings.However,these researches use only one single-shaped filter and fail to extract semantic features of different granularity.To solve this problem,ConvKG exploits multi-shaped filters to co-convolute on the triple embeddings,joint learning semantic features of different granularity.Different shaped filters cover different sizes on the triple embeddings and capture pairwise interactions of different granularity among triple elements.Experimental results confirm the strength of joint learning,and compared with state-of-the-art CNN-based KGC models,ConvKG achieves the better mean rank(MR)and Hits@10 metrics on dataset WN18 RR,and the better MR on dataset FB15k-237.
基金supported in part by the National Natural Science Foundation of China under Grant 61671074in part by Project No.A01B02C01202015D0。
文摘A distributed reinforcement learning(RL)based resource management framework is proposed for a mobile edge computing(MEC)system with both latency-sensitive and latency-insensitive services.We investigate joint optimization of both computing and radio resources to achieve efficient on-demand matches of multi-dimensional resources and diverse requirements of users.A multi-objective integer programming problem is formulated by two subproblems,i.e.,access point(AP)selection and subcarrier allocation,which can be solved jointly by our proposed distributed RL-based approach with a heuristic iteration algorithm.The proposed algorithm allows for the reduction in complexity since each user needs to consider only its own selection of AP without knowing full global information.Simulation results show that our algorithm can achieve near-optimal performance while reducing computational complexity significantly.Compared with other algorithms that only optimize either of the two sub-problems,the proposed algorithm can serve more users with much less power consumption and content delivery latency.
基金This work was supported by the Research Deanship of Prince Sattam Bin Abdulaziz University,Al-Kharj,Saudi Arabia(Grant No.2020/01/17215).Also,the author thanks Deanship of college of computer engineering and sciences for technical support provided to complete the project successfully。
文摘In the era of Big data,learning discriminant feature representation from network traffic is identified has as an invariably essential task for improving the detection ability of an intrusion detection system(IDS).Owing to the lack of accurately labeled network traffic data,many unsupervised feature representation learning models have been proposed with state-of-theart performance.Yet,these models fail to consider the classification error while learning the feature representation.Intuitively,the learnt feature representation may degrade the performance of the classification task.For the first time in the field of intrusion detection,this paper proposes an unsupervised IDS model leveraging the benefits of deep autoencoder(DAE)for learning the robust feature representation and one-class support vector machine(OCSVM)for finding the more compact decision hyperplane for intrusion detection.Specially,the proposed model defines a new unified objective function to minimize the reconstruction and classification error simultaneously.This unique contribution not only enables the model to support joint learning for feature representation and classifier training but also guides to learn the robust feature representation which can improve the discrimination ability of the classifier for intrusion detection.Three set of evaluation experiments are conducted to demonstrate the potential of the proposed model.First,the ablation evaluation on benchmark dataset,NSL-KDD validates the design decision of the proposed model.Next,the performance evaluation on recent intrusion dataset,UNSW-NB15 signifies the stable performance of the proposed model.Finally,the comparative evaluation verifies the efficacy of the proposed model against recently published state-of-the-art methods.
基金This research was supported by the BK21 FOUR(Fostering Outstanding Universities for Research)funded by the Ministry of Education(MOE,Korea)and National Research Foundation of Korea(NFR).
文摘Artificial intelligent based dialog systems are getting attention from both business and academic communities.The key parts for such intelligent chatbot systems are domain classification,intent detection,and named entity recognition.Various supervised,unsupervised,and hybrid approaches are used to detect each field.Such intelligent systems,also called natural language understanding systems analyze user requests in sequential order:domain classification,intent,and entity recognition based on the semantic rules of the classified domain.This sequential approach propagates the downstream error;i.e.,if the domain classification model fails to classify the domain,intent and entity recognition fail.Furthermore,training such intelligent system necessitates a large number of user-annotated datasets for each domain.This study proposes a single joint predictive deep neural network framework based on long short-term memory using only a small user-annotated dataset to address these issues.It investigates value added by incorporating unlabeled data from user chatting logs into multi-domain spoken language understanding systems.Systematic experimental analysis of the proposed joint frameworks,along with the semi-supervised multi-domain model,using open-source annotated and unannotated utterances shows robust improvement in the predictive performance of the proposed multi-domain intelligent chatbot over a base joint model and joint model based on adversarial learning.
基金supported by the NationalKeyR&DProgramof China underGrant No.2018YFA0701604.
文摘Due to the many types of distributed denial-of-service attacks(DDoS)attacks and the large amount of data generated,it becomes a chal-lenge to manage and apply the malicious behavior knowledge generated by DDoS attacks.We propose a malicious behavior knowledge base framework for DDoS attacks,which completes the construction and application of a multi-domain malicious behavior knowledge base.First,we collected mali-cious behavior traffic generated by five mainstream DDoS attacks.At the same time,we completed the knowledge collection mechanism through data pre-processing and dataset design.Then,we designed a malicious behavior category graph and malicious behavior structure graph for the characteristic information and spatial structure of DDoS attacks and completed the knowl-edge learning mechanism using a graph neural network model.To protect the data privacy of multiple multi-domain malicious behavior knowledge bases,we implement the knowledge-sharing mechanism based on federated learning.Finally,we store the constructed knowledge graphs,graph neural network model,and Federated model into the malicious behavior knowledge base to complete the knowledge management mechanism.The experimental results show that our proposed system architecture can effectively construct and apply the malicious behavior knowledge base,and the detection capability of multiple DDoS attacks occurring in the network reaches above 0.95,while there exists a certain anti-interference capability for data poisoning cases.
文摘中文电子病历实体关系抽取是构建医疗知识图谱,服务下游子任务的重要基础。目前,中文电子病例进行实体关系抽取仍存在因医疗文本关系复杂、实体密度大而造成医疗名词识别不准确的问题。针对这一问题,提出了基于对抗学习与多特征融合的中文电子病历实体关系联合抽取模型AMFRel(adversarial learning and multi-feature fusion for relation triple extraction),提取电子病历的文本和词性特征,得到融合词性信息的编码向量;利用编码向量联合对抗训练产生的扰动生成对抗样本,抽取句子主语;利用信息融合模块丰富文本结构特征,并根据特定的关系信息抽取出相应的宾语,得到医疗文本的三元组。采用CHIP2020关系抽取数据集和糖尿病数据集进行实验验证,结果显示:AMFRel在CHIP2020关系抽取数据集上的Precision为63.922%,Recall为57.279%,F1值为60.418%;在糖尿病数据集上的Precision、Recall和F1值分别为83.914%,67.021%和74.522%,证明了该模型的三元组抽取性能优于其他基线模型。
文摘目前,大部分将知识图谱引入推荐系统的方法只是将已知的表层知识图谱实体进行引入,没有对图谱的内在关系进行预测和挖掘,因此无法利用知识图谱中的隐藏关系。针对上述问题,提出联合学习推荐模型E-TUP(enhance towards understanding of user preference),使用E-CP(enhance canonical polyadic)进行知识图谱补全并将完整信息进行传递。利用储存空间负采样方法,将优质负例三元组进行存储,并随训练过程进行更新,以提高知识图谱补全中负例三元组的质量。链接预测实验结果显示,储存空间方法使E-TUP模型链接预测准确率对比现有模型最高提升10.3%。在MovieLens-1m和DBbook2014数据集上进行推荐实验,在多个评价指标上取得最佳结果,对比现有模型实现最高5.5%的提升,表明E-TUP可以有效利用知识图谱中的隐藏关系提高模型推荐准确率。基于汽车维修数据进行推荐实验,结果表明E-TUP可以有效推荐相关知识。