基于Stacking模型融合的用户购买行为预测研究

在大数据时代背景下,如何利用大量的销售数据精准预测顾客未来需求,成为企业制定客户管理和库存管理决策的一个重要问题。目前关于用户购买行为预测的研究中很少能够预测用户具体的购买时间。基于已有的销售数据,提出了基于机器学习和Stacking集成的综合预测模型预测用户的购买行为,即未来是否购买及其购买时间。将模型应用在一家大型连锁零售企业的需求预测中,并对方法的有效性进行评估。结果表明,基于Stacking集成的融合模型对预测用户未来是否购买具有最佳性能,准确率达85%,AUC值达到0.928;LightGBM集成算法在预测用户购买时间时具有最优性能,相比于融合模型提升了5.5%的预测性能;融合模型+LightGBM算法的组合相比于均使用融合模型提升了9.4%的预测性能。     

A positive correlation between turnout and plurality does not refute the rational voter model

Many papers have tested the prediction of the rational voter model that, ceteris paribus, turnout will be low when potential voters expect the winner's plurality to be large. The appropriate null hypothesis, however, is unclear. We show that statistical models of voting in which each voter's decision of whether to vote does not vary with the expected plurality can nonetheless generate data which lead to both positive and negative correlations between turnout and plurality.     

TACD: a transportable ant colony discrimination model for corporate bankruptcy prediction

This paper presents a transportable ant colony discrimination strategy (TACD) to predict corporate bankruptcy, a topic of vital importance that is attracting increasing interest in the field of economics. The proposed algorithm uses financial ratios to build a binary prediction model for companies with the two statuses of bankrupt and non-bankrupt. The algorithm takes advantage of an improved version of continuous ant colony optimisation (CACO) at the core, which is used to create an accurate, simple and understandable linear model for discrimination. This also enables the algorithm to work with continuous values, leading to more efficient learning and adaption by avoiding data discretisation. We conduct a comprehensive performance evaluation on three real-world data sets under a stratified cross-validation strategy. In three different scenarios, TACD is compared with 11 other bankruptcy prediction strategies. We also discuss the efficiency of the attribute selection methods used in the experiments. In addition to its simplicity and understandability, statistical significance tests prove the efficiency of TACD against the other prediction algorithms in both measures of AUC and accuracy.     

Financial Distress Prediction in an International Context: A Review and Empirical Analysis of Altman's Z‐Score Model

This paper assesses the classification performance of the Z‐Score model in predicting bankruptcy and other types of firm distress, with the goal of examining the model's usefulness for all parties, especially banks that operate internationally and need to assess the failure risk of firms. We analyze the performance of the Z‐Score model for firms from 31 European and three non‐European countries using different modifications of the original model. This study is the first to offer such a comprehensive international analysis. Except for the United States and China, the firms in the sample are primarily private, and include non‐financial companies across all industrial sectors. We use the original Z′′‐Score model developed by Altman, Corporate Financial Distress: A Complete Guide to Predicting, Avoiding, and Dealing with Bankruptcy (1983) for private and public manufacturing and non‐manufacturing firms. While there is some evidence that Z‐Score models of bankruptcy prediction have been outperformed by competing market‐based or hazard models, in other studies, Z‐Score models perform very well. Without a comprehensive international comparison, however, the results of competing models are difficult to generalize. This study offers evidence that the general Z‐Score model works reasonably well for most countries (the prediction accuracy is approximately 0.75) and classification accuracy can be improved further (above 0.90) by using country‐specific estimation that incorporates additional variables.     

Election forecasts: Cracking the Danish case

Election forecasting models based on voting theories and estimated via regression analysis are routinely available for virtually all advanced industrial democracies. Denmark, however, offers an exception, for no such prediction equations have been published on the Danish case. This absence has sometimes been attributed to the puzzling nature of economic voting there, along with the complexity of its multi-party system, which renders formulation of the dependent variable problematic. We attempt to overcome these obstacles, offering a “synthetic” forecasting model for Danish national election outcomes, 1964–2015. The regression model, based on the variables of economic growth and vote intention, performs well, by various tests. Finally, we apply it, ex ante fashion, to the 2019 contest, where the prediction favored the Social Democratic led coalition, an outcome that came to pass.     

Say-on-Pay Votes: The Role of the Media

Abstract

We investigate the association between the media coverage of firms’ CEO pay packages and subsequent shareholder voting on say-on-pay resolutions, and find that negative media coverage is able to predict shareholder discontent over say on pay. When we divide media coverage into coverage in the financial and business press versus coverage in the general press, we find that shareholder voting on say-on-pay resolutions is mainly associated with the articles from the financial and business press. This suggests that the media cannot be considered a homogeneous information source that is equally able to predict shareholders’ voting behaviors. As such, our findings have important implications for studies on the role of the media in corporate governance.     

Decreasing average cost in private schools,existence of majority voting equilibrium,and a policy analysis for Turkey

The existing models of mixed public–private school systems usually capture only the decreasing average cost faced by public schools, whereas empirical studies find evidence of it for private schools as well. Motivated by this, an equilibrium model of a mixed public–private school system is studied in this paper, whereby private schools also face decreasing average cost over enrollment. In the model, households, heterogeneous with respect to exogenously specified income and child’s ability, choose among a public and a private school. Private school charges tuition whereas public school is free. Public school spending is financed by income tax revenue collected from all households and the tax rate is determined via majority voting. Achievement of a child depends on its ability and education spending. Under the assumptions on the parameters of the model, a joint lognormal distribution of income and ability, and a Cobb–Douglas utility, majority voting equilibrium is numerically shown to exist. The model is calibrated to match certain statistics from the 2013 Turkish data. Using the calibrated model, we compare the benchmark for a mixed public–private school system with a pure public school system to understand the impact of shutting down some of the private schools in Turkey following the July 15 coup attempt. We find that mean achievement and variance of achievement after high school is \(0.039\%\) higher and \(0.013\%\) lower respectively in a pure public school system.     

Predicting the failures of prediction markets: A procedure of decision making using classification models

Prediction markets have been an important source of information for decision makers due to their high ex post accuracies. Nevertheless, recent failures of prediction markets remind us of the importance of ex ante assessments of their prediction accuracy. This paper proposes a systematic procedure for decision makers to acquire prediction models which may be used to predict the correctness of winner-take-all markets. We commence with a set of classification models and generate combined models following various rules. We also create artificial records in the training datasets to overcome the imbalanced data issue in classification problems. These models are then empirically trained and tested with a large dataset to see which may best be used to predict the failures of prediction markets. We find that no model can universally outperform others in terms of different performance measures. Despite this, we clearly demonstrate a result of capable models for decision makers based on different decision goals.     

Econometric modeling of urban and other substate areas: An analysis of alternative methodologies

Questions raised by J.R. Meyer and J.W. Millimar on the appropriate level of sophistication in regional econometric models are addressed by comparing the modeling methodologies for each of six diverse substate areas. The methodologies range from aggregated recursive structures driven by an ARIMA time-series model of export-base employment to highly detailed simultaneous equation models. It is found that recursive model accuracy is relatively insensitive to forecast accuracy of the model-driving variable and simultaneous models are more accurate than recursive ones, but relative accuracies of aggregated and detailed simultaneous models are less clear. Population and personal income estimates are improved by disaggregation, but with respect to employment, there is a trade-off in model structure between average employment prediction accuracy over a number of time periods and turning point prediction accuracy.     

Crowdsourcing-based semantic relation recognition for natural language questions over RDF data

ABSTRACT

Natural language query systems over RDF data need to rely on the semantic relations in query. First, we propose the new crowdsourcing model that used to produce semantic relations dataset. The model not only inherits completeness of the iterative model and accuracy of the parallel model, but also saves human resources. Second, we mine the rules of semantic relation recognition from the correlations between dependency structures and semantic relations. Third, we propose an algorithm of semantic relation recognition for natural language query over RDF data, and experiments demonstrate that it can recognize more semantic relations than existing methods.