利用LIME对脓毒症预测模型进行可解释性分析

针对机器学习应用于脓毒症预测存在预测准确率低和可解释性不足的问题,提出了利用LIME对基于机器学习的脓毒症预测模型进行可解释性分析.模型由预测和解释两部分组成:预测部分使用XGBoost和线性回归(LR),首先通过XGBoost进行特征提取,再利用LR对提取到的特征进行分类;解释部分使用LIME模型提取出关键的预测指标对模型进行解释.实验结果表明,通过XGBoost+LR模型进行脓毒症预测的准确率为99％,受试者工作特征曲线下面积(AUROC)为0.984,优于单独使用XGBoost(准确率:95％,AUROC:0.953)和LR(准确率:53％,AUROC:0.556)或者LGBM(准确率:90％,AUROC:0.974),同时通过LIME能有效地提取出前10个最重要的指标,对脓毒症预测模型进行可解释性分析,提高了模型的可信度.     

基于可解释集成学习的信贷违约预测

人工智能促进了风控行业的发展,智能风控的核心在于风险控制,信贷违约预测模型是解决这一问题必须倚靠的手段.传统的解决方案是基于人工和广义线性模型建立的,然而现在通过网络完成的交易数据,具有高维性和多重来源等特点,远远超出了现有模型的处理能力,对于传统风控提出了巨大的挑战.因此,本文提出一种基于融合方法的可解释信贷违约预测模型,首先选取LightGBM、DeepFM和CatBoost作为基模型,CatBoost作为次模型,通过模型融合提升预测结果的准确性,然后引入基于局部的、与模型无关的可解释性方法LIME,解释融合模型的预测结果.基于真实数据集的实验结果显示,该模型在信贷违约预测任务上具有较好的精确性和可解释性.     

深度学习可解释性内涵和分类

近年来,深度学习在多个行业得到了广泛应用,效果显著。深度学习虽然具有数学统计原理基础,但是对于任务知识表征学习尚缺乏明确解释。对深度学习理论研究的缺乏将导致即时可以通过各种训练方法使得模型得到满意输出,但不能解释模型内部究竟如何进行工作才得到有效结果。本文从深度学习可解释性内涵和分类角度出发,阐述了深度学习可解释性,以期有助于其他学者研究。     

目标检测模型的决策依据与可信度分析

目标检测模型已经在很多领域得到广泛应用, 但是, 作为一种机器学习模型, 对人类来说仍然是一个黑盒. 对模型进行解释有助于我们更好地理解模型, 并判断其可信度. 针对目标检测模型的可解释性问题, 提出将其输出改造为关注每一类物体存在性概率的具体回归问题, 进而提出分析目标检测模型决策依据与可信度的方法. 由于原有图像分割方法的泛用性较差, 解释目标检测模型时, LIME所生成解释的忠诚度较低、有效特征数量较少. 提出使用DeepLab代替LIME的图像分割方法, 以对其进行改进. 改进后的方法可以适用于解释目标检测模型. 实验的对比结果证明了所提出改进方法在解释目标检测模型时的优越性.     

一种即时软件缺陷预测模型及其可解释性研究

即时软件缺陷预测是保障软件安全与质量相统一的必要途径,在软件工程领域受到越来越多的关注.然而,现有数据集存在特征冗余和特征相关性低的情况,极大影响了即时软件缺陷预测模型的分类性能和稳定性.此外,分析缺陷数据特征对模型的影响尤为重要,但如今对软件缺陷预测模型进行解释性研究较少.针对这些问题,文章基于6个开源项目的2274...     

基于可解释性的Android恶意软件检测

针对Android恶意软件检测,通常仅有检测结果缺乏对其检测结果的可解释性.基于此,从可解释性的角度分析Android恶意软件检测,综合利用多层感知机和注意力机制提出一种可解释性的Android恶意软件检测方法(multilayer perceptron attention-method, MLP＿At).通过提取Android恶意软件的应用权限和应用程序接口(application programming interface, API)特征来进行数据预处理生成特征信息,采用多层感知机对特征学习.最后,利用BP算法对学习到的数据进行分类识别.在多层感知机中引入注意力机制,以捕获敏感特征,根据敏感特征生成描述来解释应用的核心恶意行为.实验结果表明所提方法能有效检测恶意软件,与SVM、RF、XGBoost相比准确率分别提高了3.65%、3.70%和2.93%,并能准确地揭示软件的恶意行为.此外,该方法还可以解释样本被错误分类的原因.     

基于Android操作系统的移动终端App开发研究

随着市场的不断需求,Android操作系统越来越完善,由于时代的发展需要,App也逐渐成为市场空缺,App终端研究成为主流研究项目.该文对基于Android操作系统的移动终端App开发视角下进行深入细致的剖析,为相关领域从业人员提供一定理论指导和参考,更好地促进了在当前互联网不断发展普及背景下Android操作系统的移...     

基于机器学习的Android App违规收集个人信息行为检测方法研究

随着移动互联网的迅猛发展,我国App数量和用户规模激增,极大便利了用户的生产生活。但另一方面,App个人信息收集过度现象普遍,因此加强App所涉个人信息保护迫在眉睫。本文提出一种基于机器学习的Android App违规收集个人信息行为的检测方法,对App违规行为进行评估,能够有效地对Android App合规性进行检测,节约人工成本,提升检测效率。     

即时软件缺陷预测研究进展

软件缺陷预测一直是软件工程研究中最活跃的领域之一,研究人员己经提出了大量的缺陷预测技术,根据预测粒度不同,主要包括模块级、文件级和变更级（change-level）缺陷预测.其中,变更级缺陷预测旨在于开发者提交代码时,对其引入的代码是否存在缺陷进行预测,因此又被称作即时（just-in-time）缺陷预测.近年来,即时缺陷预测技术由于其即时性、细粒度等优势,成为缺陷预测领域的研究热点,取得了一系列研究成果;同时也在数据标注、特征提取、模型评估等环节面临诸多挑战,迫切需要更先进、统一的理论指导和技术支撑.鉴于此,从即时缺陷预测技术的数据标注、特征提取和模型评估等方面对近年来即时缺陷预测研究进展进行梳理和总结.主要内容包括：（1）归类并梳理了即时缺陷预测模型构建中数据标注常用方法及其优缺点;（2）对即时缺陷预测的特征类型和计算方法进行了详细分类和总结;（3）总结并归类现有模型构建技术;（4）总结了模型评估中使用的实验验证方法与性能评估指标;（5）归纳出了即时缺陷预测技术的关键问题;（6）最后展望了即时缺陷预测的未来发展.     

基于神经回路策略的无线网络流量可解释性预测方法

针对无线网络流量预测任务中模型的“黑箱”问题,提出基于神经回路策略（NeuralCircuitPolicy,NCP）的高可解释性流量预测模型。首先,采用均值滤波分解流量数据的趋势、残差分量来增强特征表示;其次,构建紧凑的NCP框架以可信学习策略捕捉流量长期时变关系;最后,通过全连接层得到预测结果。实验表明,该方法以可解释的方式实现了精准的无线网络流量预测。     

Multi-task deep neural networks for just-in-time software defect prediction on mobile apps

With the development of smartphones, mobile applications play an irreplaceable role in our daily life, which characteristics often commit code changes to meet new requirements. This characteristic can introduce defects into the software. To provide immediate feedback to developers, previous researchers began to focus on just-in-time (JIT) software defect prediction techniques. JIT defect prediction aims to determine whether code commits will introduce defects into the software. It contains two scenarios, within-project JIT defect prediction and cross-project JIT defect prediction. Regardless of whether within-project JIT defect prediction or cross-project JIT defect prediction all need to have enough labeled data (within-project JIT defect prediction assumes that have plenty of labeled data from the same project, while cross-project JIT defect prediction assumes that have sufficient labeled data from source projects). However, in practice, both the source and target projects may only have limited labeled data. We propose the MTL-DNN method based on multi-task learning to solve this question. This method contains the data preprocessing layer, input layer, shared layers, task-specific layers, and output layer. Where the common features of multiple related tasks are learned by sharing layers, and the unique features of each task are learned by the task-specific layers. For verifying the effectiveness of the MTL-DNN approach, we evaluate our method on 15 Android mobile apps. The experimental results show that our method significantly outperforms the state-of-the-art single-task deep learning and classical machine learning methods. This result shows that the MTL-DNN method can effectively solve the problem of insufficient labeled training data for source and target projects.     

Effort-aware cross-project just-in-time defect prediction framework for mobile apps

As the boom of mobile devices, Android mobile apps play an irreplaceable roles in people’s daily life, which have the characteristics of frequent updates involving in many code commits to meet new requirements. Just-in-Time (JIT) defect prediction aims to identify whether the commit instances will bring defects into the new release of apps and provides immediate feedback to developers, which is more suitable to mobile apps. As the within-app defect prediction needs sufficient historical data to label the commit instances, which is inadequate in practice, one alternative method is to use the cross-project model. In this work, we propose a novel method, called KAL, for cross-project JIT defect prediction task in the context of Android mobile apps. More specifically, KAL first transforms the commit instances into a high-dimensional feature space using kernel-based principal component analysis technique to obtain the representative features. Then, the adversarial learning technique is used to extract the common feature embedding for the model building. We conduct experiments on 14 Android mobile apps and employ four effort-aware indicators for performance evaluation. The results on 182 cross-project pairs demonstrate that our proposed KAL method obtains better performance than 20 comparative methods.     

基于代码自然性的切片粒度缺陷预测方法

软件缺陷预测是软件质量保障领域的一个活跃话题,它可以帮助开发人员发现潜在的缺陷并更好地利用资源.如何为预测系统设计更具判别力的度量元,并兼顾性能与可解释性,一直是人们致力于的研究方向.针对这一挑战,提出了一种基于代码自然性特征的缺陷预测方法——CNDePor.该方法通过正逆双向度量代码和利用质量信息对样本加权的方式改进语言模型,提高了模型所得交叉熵（CE）类度量元的缺陷判别力.针对粗粒度缺陷预测存在难以聚焦缺陷区域、代码审查成本高的不足,研究了一种新的细粒度缺陷预测问题——面向语句的切片级缺陷预测.在此问题上,设计了4种度量元,并在两类安全缺陷数据集上验证了度量元和CNDePor方法的有效性.实验结果表明：CE类度量元具有可学习性,它们蕴涵了语言模型从语料库中学习到的相关知识;改进的CE类度量元的判别力明显优于原始度量元和传统规模度量元;CNDePor方法较传统缺陷预测方法和已有的基于代码自然性的方法有显著优势,较先进的基于深度学习的方法具有可比性性能和更强的可解释性.     

An explainable deep learning model for prediction of early-stage chronic kidney disease

Chronic kidney disease (CKD) is a major public health concern with rising prevalence and huge costs associated with dialysis and transplantation. Early prediction of CKD can reduce the patient's risk of CKD progression to end-stage kidney failure. Artificial intelligence offers more intelligent and expert healthcare services in disease diagnosis. In this work, a deep learning model is built using deep neural networks (DNN) with an adaptive moment estimation optimization function to predict early-stage CKD. The health care applications require interpretability over the predictions of the black-box model to build conviction towards the model's prediction. Hence, the predictions of the DNN-CKD model are explained by the local interpretable model-agnostic explainer (LIME). The diagnostic patient data is trained on five layered DNN with three hidden layers. Over the unseen data, the DNN-CKD model yields an accuracy of 98.75% and a roc_auc score of 98.86% in detecting CKD risk. The explanation revealed by the LIME algorithm echoes the influence of each feature on the prediction made by the DNN-CKD model over the given CKD data. With its interpretability and accuracy, the proposed system may effectively help medical experts in the early diagnosis of CKD.     

基于出行模式的注意力机制可解释性探索

为了探索深度注意力模型在地铁出行预测任务中的可解释性,提出基于出行模式的注意力权重擦除方法和可解释性评估框架。利用提出的地铁出行深度注意力框架搭建预测模型,使用广州地铁羊城通数据构造三种不同长度出行序列数据集进行模型训练和验证,达到70%以上准确率;通过单一出行模式的注意力权重擦除实验发现,擦除最大注意力权重的出行模式比随机模式更能显著地影响模型预测结果,但大多数样本不发生预测结果的变化。即注意力机制在该条件下提供的可解释性信息是有限的,且该信息量随着序列长度增加而减小;通过一组出行模式注意力权重擦除实验结果表明,按注意力权重降序擦除能最快使模型预测结果发生变化,并且模型能稳定地对重要的出行模式的出行记录分配注意力权重,即注意力机制在该条件下较好地提供了可解释性信息,且该信息量随着序列长度增加而增大。     

Studying just-in-time defect prediction using cross-project models

Unlike traditional defect prediction models that identify defect-prone modules, Just-In-Time (JIT) defect prediction models identify defect-inducing changes. As such, JIT defect models can provide earlier feedback for developers, while design decisions are still fresh in their minds. Unfortunately, similar to traditional defect models, JIT models require a large amount of training data, which is not available when projects are in initial development phases. To address this limitation in traditional defect prediction, prior work has proposed cross-project models, i.e., models learned from other projects with sufficient history. However, cross-project models have not yet been explored in the context of JIT prediction. Therefore, in this study, we empirically evaluate the performance of JIT models in a cross-project context. Through an empirical study on 11 open source projects, we find that while JIT models rarely perform well in a cross-project context, their performance tends to improve when using approaches that: (1) select models trained using other projects that are similar to the testing project, (2) combine the data of several other projects to produce a larger pool of training data, and (3) combine the models of several other projects to produce an ensemble model. Our findings empirically confirm that JIT models learned using other projects are a viable solution for projects with limited historical data. However, JIT models tend to perform best in a cross-project context when the data used to learn them are carefully selected.     

Investigating decision factors in mobile application purchase: A mixed-methods approach

The growth of the smart devices market and the development of mobile applications (Apps) for them have given rise to an App economy. Sales of mobile applications are a key revenue source in this economy, with the expected worldwide market growth of US $75 billion by 2017. Despite the trend, many mobile Apps fail to attract customers, yet there has been a lack of research and understanding of the factors that affect the decisions to buy them. This study is thus motivated to examine the factors that people consider in their buying decisions of mobile Apps for their smartphones. This mixed-methods investigation first adopts an exploratory and qualitative approach to identify the purchase decision factors based on interviews with consumers. It then undertakes a quantitative, confirmatory study using a survey to test the model derived using mental accounting theory and the findings of the exploratory study. The results show the direct and indirect effects of five factors – word of mouth about App, App usefulness, monetary value of App, App trialability, and App enjoyment – on the intention to purchase an App. In this manner, this study advances our understanding of the decision-making factors leading to the purchase of mobile Apps. It also facilitates developers and marketers to promote the sales of their Apps for revenue generation.     

基于Android系统的移动医疗项目开发

移动医疗是通过使用移动电话、卫星通信等移动通信技术来提供医疗信息和服务,具体到移动互联网领域,则以基于Android和iOS等移动终端系统的医疗健康类App应用为主。为提高医疗App应用的交互性,利用关系数据库基于Android系统开发了一款用于疾病自诊的应用。主要介绍了疾病自诊中关系数据库的设计和开发过程中用到的编程技术。测试结果表明,此应用交互性好、可靠性高,为发展中国的医疗卫生服务提供了一种有效方法。