A Probabilistic Rating Prediction and Explanation Inference Model for Recommender Systems

Collaborative Filtering(CF) is a leading approach to build recommender systems which has gained considerable development and popularity. A predominant approach to CF is rating prediction recommender algorithm, aiming to predict a user's rating for those items which were not rated yet by the user. However, with the increasing number of items and users, thedata is sparse.It is difficult to detectlatent closely relation among the items or users for predicting the user behaviors. In this paper,we enhance the rating prediction approach leading to substantial improvement of prediction accuracy by categorizing according to the genres of movies. Then the probabilities that users are interested in the genres are computed to integrate the prediction of each genre cluster. A novel probabilistic approach based on the sentiment analysis of the user reviews is also proposed to give intuitional explanations of why an item is recommended.To test the novel recommendation approach, a new corpus of user reviews on movies obtained from the Internet Movies Database(IMDB) has been generated. Experimental results show that the proposed framework is effective and achieves a better prediction performance.

Analysis on Explanation Effect of the European Numerical Prediction on Temperature

[Objective] The research aimed to analyze explanation effect of the European numerical prediction on temperature. [Method] Based on CMSVM regression method, by using 850 hPa grid point data of the European numerical prediction from 2003 to 2009 and actual data of the maximum and minimum temperatures at 8 automatic stations in Qingyang City, prediction model of the temperature was established, and running effect of the business from 2008 to 2010 was tested and evaluated. [Result] The method had very good guidance role in real-time business running of the temperature prediction. Test and evaluation found that as forecast time prolonged, prediction accuracies of the maximum and minimum temperatures declined. When temperature anomaly was higher (actual temperature was higher than historical mean), prediction accuracy increased. Influence of the European numerical prediction was bigger. [Conclusion] Compared with other methods, operation of the prediction method was convenient, modeling was automatic, running time was short, system was stable, and prediction accuracy was high. It was suitable for implementing of the explanation work for numerical prediction product at meteorological station.

Joint Probability Prediction Model of Rainfall Triggered Landslides and Debris Flows

The rainfall induced landslides and debris flows are the major disasters in China, as well in Europe, South America, Japan and Australia. This paper proposes a new type of joint probability prediction model—Double Layer Nested Multivariate Compound Extreme Value Distribution (DLNMCEVD) to predict landslides and debris flows triggered by rainfall. The outer layer of DLNMCEVD is predicting the joint probabilities of different combinations for rainfall characteristics, air temperature and humidity, which should be considered as external load factors with geological and geotechnical characteristics as resistance factors for reliability analysis of slope stability in the inner layer of model. For the reliability and consequence analysis of rainfall-induced slope failure, the Global Uncertainty Analysis and Global Sensitivity Analysis (GUA & GSA) should be taken into account for input-output iterations. Finally, based on the statistics prediction by DLNMCEVD, the geological hazards prevention alarm and regionalization can be provided in this paper.

Comparison of Sales Prediction in Conventional Insights and Machine Learning Perspective

With the integration of global economy development and the rapid growth of science knowledge and technology,the needs of people’s consumption are increasingly personalized and diversified.Such a market background makes sales forecasting become an indispensable part of enterprise management and development.The definition of the sales forecasting is that based on the past few years’sales situation,the enterprises through systematic sales forecasting models estimate of the quantity and amount of all or some specific sales products and services in a specific time in the future.Accurate sales forecasting can promote enterprises to do better in future revenue,and can also encourage enterprises to set and keep an efficient sales management team.This paper will analyze traditional sales forecasting methods and sales forecasting methods based on big data models related to the perspective of machine learning,and then compare them.The research shows that the two sales forecasting methods have their own advantages and disadvantages.In the future,enterprises can adopt the two sales forecasting methods in parallel to maximize the utilization advantage of sales forecasting for enterprises.

Prediction of Kappa Number in Eucalyptus Kraft Pulp Continuous Digester Using the Box &Jenkins Methodology

The quality of the resulting pulping continuous digesters is monitored by measuring the Kappa number, which is a reference of residual lignin. The control of the kappa number is carried out mainly in the top of the digester, therefore it is important to get some indication of this analysis beforehand. In this context, the aim of this work was to obtain a prediction model of the kappa number in advance to the laboratory results. This paper proposes a new approach using the Box & Jenkins methodology to develop a dynamic model for predicting the kappa number from a Kamyr continuous digester from an eucalyptus Kraft pulp mill in Brazil. With a database of 1500 observations over a period of 30 days of operation, some ARMA models were studied, leading to the choice of ARMA (1, 2) as the best forecasting model. After fitting the model, we performed validation with a new set of data from 30 days of operation, achieving a model of 2.7% mean absolute percent error.

Prediction of Favorable Gas Accumulation Zones in the Western Part of Southern Margin in Junggar Basin

ＰｒｅｄｉｃｔｉｏｎｏｆＦａｖｏｒａｂｌｅＧａｓＡｃｃｕｍｕｌａｔｉｏｎＺｏｎｅｓｉｎｔｈｅＷｅｓｔｅｒｎＰａｒｔｏｆＳｏｕｔｈｅｒｎＭａｒｇｉｎｉｎＪｕｎｇｇａｒＢａｓｉｎ￥ＳｏｎｇＹａｎ，ＤａｉＪｉｎｘｉｎｇ，ＷａｎｇＤａｒｕｉ，ＳｕｎＹｏｎｇｘ...

Explainable prediction of loan default based on machine learning models

Owing to the convenience of online loans,an increasing number of people are borrowing money on online platforms.With the emergence of machine learning technology,predicting loan defaults has become a popular topic.However,machine learning models have a black-box problem that cannot be disregarded.To make the prediction model rules more understandable and thereby increase the user’s faith in the model,an explanatory model must be used.Logistic regression,decision tree,XGBoost,and LightGBM models are employed to predict a loan default.The prediction results show that LightGBM and XGBoost outperform logistic regression and decision tree models in terms of the predictive ability.The area under curve for LightGBM is 0.7213.The accuracies of LightGBM and XGBoost exceed 0.8.The precisions of LightGBM and XGBoost exceed 0.55.Simultaneously,we employed the local interpretable model-agnostic explanations approach to undertake an explainable analysis of the prediction findings.The results show that factors such as the loan term,loan grade,credit rating,and loan amount affect the predicted outcomes.

山区公路交叉口驾驶避险决策行为特性分析

为研究驾驶特征指标与驾驶避险行为决策的关联规则以提升驾驶安全,将驾驶避险决策行为划分为纵向“停车避险”和横向“转向避险”,并运用驾驶模拟系统构建12种山区公路交叉口交通冲突虚拟场景,招募38名驾驶人进行试验,采集车辆特征和驾驶人扫视、眨眼、注视等眼动特征数据。基于随机森林构建驾驶避险决策行为模型,然后引入沙普利加和解释(SHapley Additive exPlanation, SHAP)框架进一步分析车辆、眼动特征与驾驶避险行为之间的非线性关系。结果表明:模型对纵、横向避险行为预测的准确率分别为84.77%、94.70%;纵向速度标准差、扫视持续时间标准差、轨迹偏差标准差、侧向速度标准差与驾驶避险决策行为存在明显关联,如纵向速度标准差过大(约大于10 km/h),纵向“停车避险”可能性明显增加。

基于HA-RF-SHAP的露天煤矿粉尘浓度预测模型

为了有效预测和控制煤矿粉尘浓度,保障煤矿工人健康及环境安全,以宝日希勒露天煤矿现场粉尘监测数据为基础,使用随机森林对粉尘浓度进行预测,提出了4种启发式智能优化算法优化随机森林超参数的方法,通过RMSE、MAE和皮尔逊相关系数R对模型进行评价,采用SHAP可解释模型分析影响露天煤矿粉尘浓度的因素。结果表明:PM2.5、PM10、TSP的最优模型分别为GWO-RF、WOA-RF和HHO-RF;超参数调整使模型整体RMSE指标提升约为1~3,MAE提升约为1~2.5,R提升约4%~6%;PM2.5的预测表现最好,训练集与测试集共同作用时,R为0.9463,MAE为3.059,RMSE为4.919,其次是PM10、TSP;单因素作用时,湿度对于该矿粉尘浓度影响最大,双因素同时影响下湿度和气压对粉尘浓度变化影响最大。研究提供了一个有效的粉尘浓度预测方法,可准确预测粉尘浓度并确定粉尘最影响因素,对矿山粉尘管控具有重要参考价值。

基于北向资金动向的股指收益率预测研究

自“沪港通”和“深港通”开通以来,北向投资者的交易行为始终被市场所紧密地关注,对于未来的市场走势具有重要的引领作用。本文构建了反映北向投资者资金动向的代理指标(NCM),并且采用样本内的回归预测模型和样本外的R~2统计量以及全包围测试方法,研究了其对于股票市场收益率的预测能力。实证结果表明,北向资金的动向能够有效地预测2,5,10和20个交易日的股票市场收益率,这种预测能力对格兰杰因果检验及不同的资金动向计量方法和检验样本都具有较好的稳健性。进一步研究发现,NCM的预测能力主要来自参与北向交易的投资者分析和处理未来现金信息的能力。最后,利用这一预测能力构建量化择时策略,发现NCM市场择时策略可以在2,5,10和20个交易日的持有期内具有较好的投资绩效表现。

基于卷积神经网络的液化预测模型及可解释性分析

常规液化判别方法通常是半经验方法,存在人为因素干扰,成功率及均衡性不佳。现有的机器学习方法缺乏足够的样本支撑,存在一定的局限性。通过整合液化数据集,选取修正标准贯击数、细粒含量、土层深度、地下水位深度、总上覆应力、有效上覆应力、门槛加速度、循环剪应力比、剪切波速、震级与地表峰值加速度11个液化特征建立卷积神经网络(convolutional neural network,简称CNN)模型。引入边界合成少数过采样技术消除不平衡数据集的影响。将CNN模型与随机森林模型、逻辑回归模型、支持向量机模型、极致梯度提升模型和规范方法进行对比,并结合沙普利加性解释(SHapley Additive exPlanations,简称SHAP)分析输入特征对预测结果的影响趋势。结果表明,CNN模型准确率达92.58%,各项指标均优于其他4种机器学习模型和规范方法。对SHAP结果分析可知,修正标贯击数小于15的土层液化概率较高,循环剪应力比CSR小于0.25的土层更不易液化。各因素的影响规律均符合现有认知,预测模型合理可靠。

基于LSTM-多头混合注意力的可解释换道意图预测

为了使自动驾驶汽车准确地预测其周围车辆的换道意图,提出了一种基于长短期记忆神经网络(LSTM)-多头混合注意力的可解释换道意图预测模型。该模型可以充分提取目标车辆与其周围车辆之间的时空交互关系,并且提出了一种基于最大熵的Shapley加性解释方法(SHAP)来解释各个特征在特定时间步对模型输出的影响程度,在HighD数据集上进行了实验。并通过SHAP值的可视化,直观解释了换道预测模型在特定时刻的目标车辆的换道行为。结果表明:该换道预测模型在换道前3 s的综合准确率,分别比LSTM、卷积神经网络(CNN)、多头注意力高出4.03%、9.51%、5.16%,这证明了模型在长时域预测的有效性;错误预测样本归因于模型缺陷或样本稀疏。该换道预测模型可为用户进行模型优化提供指导。

基于校准窗口集成与耦合市场特征的可解释双层日前电价预测

随着电力市场之间耦合程度不断加深,只局限于单个市场内部的传统特征集不足以支撑高精度预测的需求。而且模型预测性能对校准窗口的选择敏感,而传统电价预测仅使用一个固定时间长度的数据集,同时预测模型的“黑盒”结构导致预测结果在工程应用中可信度偏低。针对上述问题,该文提出一种考虑校准窗口集成与耦合市场特征的可解释双层日前电价预测框架。内层框架为基于改进自适应噪声完备集合经验模态分解(improved complete ensemble empirical mode decomposition,ICEEMDAN)的择优预测,首先分解原始电价序列,然后应用Lasso估计回归(lassoestimated autoregressive,LEAR)、长期和短期时间序列网络(long-term and short-term time-series networks,LSTNet)、卷积神经网络-长短记忆神经网络(convolutionalneuralnetworks-longshort termmemory,CNN-LSTM)、移动平均(autoregressive integrated moving average,ARIMA)和核极限学习机(kernel extreme learning machines,KELM)模型预测子序列并选择最优预测算法。外层框架为基于贝叶斯模型平均(bayes modelaveraging,BMA)的校准窗口集成预测,针对每个不同校准窗口长度数据集下的预测分配权重并集成得到预测电价。最后,通过可解释方法沙普利加性解释模型(shapley additiveexplanations,SHAP)分析耦合市场特征如何影响预测电价。该文通过北欧电力市场数据集的算例分析证明了所提算法的优越性和校准窗口集成方案的有效性。

基于机器学习构建急性心力衰竭患者易损期死亡或再入院风险预测模型

目的应用不同机器学习算法构建急性心力衰竭(acute heart failure,AHF)患者易损期死亡或再入院的风险预测模型,并筛选出最优模型。方法选取2019年10月至2021年7月于陆军军医大学第二附属医院心血管内科住院治疗的651例AHF患者为研究对象,收集入院生命体征、合并症和实验室检查结果等临床资料。复合终点事件定义为AHF患者出院后3个月内发生全因死亡或心衰加重再入院。采用简单随机抽样法将研究对象按8∶2拆分为训练集(521例)和测试集(130例),基于逻辑回归(logistic regression,LR)、随机森林(random forest,RF)、决策树(decision tree,DT)、轻量梯度提升(light gradient boosting machine,LGBM)、极端梯度提升(extreme gradient boosting,XGBoost)和神经网络(neural network,NN)6种机器学习算法分别构建预测模型。采用受试者工作特征曲线(receiver operating characteristic,ROC)和临床决策曲线(decision curve analysis,DCA)对模型的预测性能和临床获益进行评价,使用Shapley加性解释(Shapley additive explanation,SHAP)算法评估不同临床特征对模型的影响。结果651例AHF患者中发生复合终点事件203例(31.2%)。ROC曲线分析显示,LR、RF、DT、LGBM、XGBoost和NN模型的曲线下面积(area under curve,AUC)依次为0.707、0.756、0.616、0.677、0.768、0.681,XGBoost模型的AUC最高,DCA曲线中XGBoost模型的临床决策净获益也更大,整体预测效能最佳。SHAP算法分析得出,影响XGBoost模型输出结果的重要临床特征分别为血清尿酸、D-二聚体、平均动脉压、B型利钠肽、左房前后径、体质量指数和NYHA分级。结论XGBoost模型预测急性心力衰竭患者易损期死亡或再入院风险效果最佳。

基于贝叶斯方法的四川暴雨集合概率预报产品释用

在贝叶斯概率决策理论的基础上,探索了一种提高四川暴雨预报准确率的方法,该方法利用四川境内1951—2004年147站暴雨的气候概率对西南区域中尺度集合预报模式提供的≥50mm集合降水概率预报产品进行了修正。从2008年6—9月的连续性预报试验结果来看:基于贝叶斯方法修正后的集合概率预报产品在一定程度上消除了空报,尽管相比于区域中尺度集合预报系统直接提供的暴雨概率预报产品实际提高暴雨准确率的效果不明显,但在如何有效地利用数值集合预报产品提高四川暴雨预报的准确率以及如何为预报员提供更有价值的暴雨预警决策方法上值得进一步探索。

归类不确定情景下特征推理的综合条件概率模型

用大学生被试,通过三个实验探讨在集中呈现类别成员样本信息的归类不确定情景下的特征推理。实验结果表明,单纯的归类确定性程度和靶类别靶特征的代表性并不直接影响被试的特征推理,而是预测特征相对于目标特征的综合条件概率直接影响被试的特征推理;特征推理不是基于类别中介的间接推理,而是基于特征关联综合条件概率的直接推理。实验结果支持作者提出的预测特征综合条件概率模型。

汉语历史语法研究的目标

本文认为,汉语历史语法研究的目标是揭示已有演变的规律、解释共时语言现象、预测未来演变的方向。此外,本文还尝试提出汉语历史语法研究中几个亟需开展或加强的研究课题。

双耦合时效曲线与解析判据滑坡预报实例分析

在某高边坡滑坡预报过程中,基于岩土结构稳定性基础理论以及双耦合时效曲线与解析的应用,描述了岩土体滑坡从流变萌发、蠕变发育到失稳剧滑的时间轨迹,使得滑坡发生的时间、规模和趋势得到精度预测、预报。在第一时间判断险情并采取险情处置措施,确保了施工安全,为既有矿坑高边坡的滑坡精确预测、预报、安全施工提供了一定的借鉴与参考。

一个创业理论的诞生——面向实践的管理研究之内涵与成果评价

通过分析《创业行动逻辑论》的产生过程,阐述了面向实践的管理研究之内涵,以明确立场,建立共识,避免不必要的误解;然后,根据学术传统,提出评价面向实践的管理研究成果,可以采用学术成果评价的共同标准,即:解决问题的重要性、研究方法的严谨性、解释与预见现象的有效性三大评价指标,不必另外为其设置评价指标体系;最后,根据研究成果的评价指标,预示了面向实践的管理研究未来努力的方向。

新市场财政学:旨在增强财政学解释力的新范式

"公共财政学"作为当代主流财政理论存在的最为突出的问题就是解释力和预测力严重弱化。"新市场财政学"就是针对当代主流财政理论的这个问题所构建的一个新范式。在论文中,笔者主要是运用规范研究方法,以政府和市场之间的关系作为切入点,深入剖析了当代主流财政理论存在的致命缺陷,阐述了新市场财政学通过重新定义市场模型,重新解释政府(以及以政府为代表的公共部门)、企业(以及以企业为代表的私人部门)和市场之间的关系,进而为重构具有强大解释力和精准预测力的财政学范式开辟了一个全新的研究路径。论文以对我国改革开放以来财政理论建设与发展的历程为线索介绍了新市场财政学范式形成的简要过程,同时从核心概念体系、理论渊源以及理论与实践的关系等三个维度描述了新市场财政学范式的基本特征,初步阐释了探索构建新市场财政学范式的目的是要增强财政科学的解释力和预测力这一财政基础理论研究的宗旨。