Health insurance fraud increases the inefficiency and inequality in our society. To address the widespread problem, cost effect techniques are in need to detect fraudulent claims. With a dataset from medical expense insurance in China, we propose a discrete choice model to identify predicting factors of fraudulent claims, and we address the major limitations of discrete choice model by considering over sampling of fraudulent cases, as well as mislabeling of legitimate claims (omission error). Our results show that a few factors, such as hospital’s qualification and policyholder’s renewal status, could be used to predict fraudulent claims for further investigation.
{"title":"Detection of Health Insurance Fraud with Discrete Choice Model: Evidence from Medical Expense Insurance in China","authors":"Yi Yao, Qixiang Sun, Shan-Hui Lin","doi":"10.2139/ssrn.2459343","DOIUrl":"https://doi.org/10.2139/ssrn.2459343","url":null,"abstract":"Health insurance fraud increases the inefficiency and inequality in our society. To address the widespread problem, cost effect techniques are in need to detect fraudulent claims. With a dataset from medical expense insurance in China, we propose a discrete choice model to identify predicting factors of fraudulent claims, and we address the major limitations of discrete choice model by considering over sampling of fraudulent cases, as well as mislabeling of legitimate claims (omission error). Our results show that a few factors, such as hospital’s qualification and policyholder’s renewal status, could be used to predict fraudulent claims for further investigation.","PeriodicalId":165362,"journal":{"name":"ERN: Discrete Regression & Qualitative Choice Models (Single) (Topic)","volume":"46 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2014-06-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124630216","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Marketing variables that are included in consumer discrete choice models are often endogenous. Extant treatments using likelihood-based estimators impose parametric distributional assumptions, such as normality, on the source of endogeneity. These assumptions are restrictive because misspecified distributions have an impact on parameter estimates and associated elasticities. The normality assumption for endogeneity can be inconsistent with some marginal cost specifications given a price-setting process, although they are consistent with other specifications. In this paper, we propose a heterogeneous Bayesian semiparametric approach for modeling choice endogeneity that offers a flexible and robust alternative to parametric methods. Specifically, we construct centered Dirichlet process mixtures CDPM to allow uncertainty over the distribution of endogeneity errors. In a similar vein, we also model consumer preference heterogeneity nonparametrically via a CDPM. Results on simulated data show that incorrect distributional assumptions can lead to poor recovery of model parameters and price elasticities, whereas the proposed semiparametric model is able to robustly recover the true parameters in an efficient fashion. In addition, the CDPM offers the benefits of automatically inferring the number of mixture components that are appropriate for a given data set and is able to reconstruct the shape of the underlying distributions for endogeneity and heterogeneity errors. We apply our approach to two scanner panel data sets. Model comparison statistics indicate the superiority of the semiparametric specification and the results show that parameter and elasticity estimates are sensitive to the choice of distributional forms. Moreover, the CDPM specification yields evidence of multimodality, skewness, and outlying observations in these real data sets. Data, as supplemental material, are available at http://dx.doi.org/10.1287/mnsc.2013.1811 . This paper was accepted by J. Miguel Villas-Boas, marketing.
包含在消费者离散选择模型中的营销变量通常是内生的。现有的基于似然估计的处理方法对内生性的来源施加了参数分布假设，如正态性。这些假设是限制性的，因为错误指定的分布对参数估计和相关的弹性有影响。内生性的正态性假设可能与给定定价过程的某些边际成本规格不一致，尽管它们与其他规格一致。在本文中，我们提出了一种异构贝叶斯半参数方法来建模选择内生性，它提供了一种灵活和鲁棒的替代参数方法。具体来说，我们构建了中心狄利克雷过程混合CDPM，以允许内质性误差分布的不确定性。同样，我们也通过CDPM对消费者偏好异质性进行非参数化建模。模拟数据的结果表明，不正确的分布假设会导致模型参数和价格弹性的较差恢复，而所提出的半参数模型能够有效地鲁棒恢复真实参数。此外，CDPM还提供了自动推断适合给定数据集的混合成分数量的好处，并且能够重建内质性和异质性误差的潜在分布形状。我们将我们的方法应用于两个扫描面板数据集。模型比较统计表明了半参数规范的优越性，结果表明参数估计和弹性估计对分布形式的选择很敏感。此外，CDPM规范在这些真实数据集中产生了多模态、偏态和离群观测的证据。作为补充资料的数据可在http://dx.doi.org/10.1287/mnsc.2013.1811上获得。这篇论文被市场营销学的J. Miguel Villas-Boas接受。
{"title":"A Bayesian Semiparametric Approach for Endogeneity and Heterogeneity in Choice Models","authors":"Yang Li, Asim Ansari","doi":"10.2139/ssrn.2270993","DOIUrl":"https://doi.org/10.2139/ssrn.2270993","url":null,"abstract":"Marketing variables that are included in consumer discrete choice models are often endogenous. Extant treatments using likelihood-based estimators impose parametric distributional assumptions, such as normality, on the source of endogeneity. These assumptions are restrictive because misspecified distributions have an impact on parameter estimates and associated elasticities. The normality assumption for endogeneity can be inconsistent with some marginal cost specifications given a price-setting process, although they are consistent with other specifications. In this paper, we propose a heterogeneous Bayesian semiparametric approach for modeling choice endogeneity that offers a flexible and robust alternative to parametric methods. Specifically, we construct centered Dirichlet process mixtures CDPM to allow uncertainty over the distribution of endogeneity errors. In a similar vein, we also model consumer preference heterogeneity nonparametrically via a CDPM. Results on simulated data show that incorrect distributional assumptions can lead to poor recovery of model parameters and price elasticities, whereas the proposed semiparametric model is able to robustly recover the true parameters in an efficient fashion. In addition, the CDPM offers the benefits of automatically inferring the number of mixture components that are appropriate for a given data set and is able to reconstruct the shape of the underlying distributions for endogeneity and heterogeneity errors. We apply our approach to two scanner panel data sets. Model comparison statistics indicate the superiority of the semiparametric specification and the results show that parameter and elasticity estimates are sensitive to the choice of distributional forms. Moreover, the CDPM specification yields evidence of multimodality, skewness, and outlying observations in these real data sets. \u0000 \u0000Data, as supplemental material, are available at http://dx.doi.org/10.1287/mnsc.2013.1811 . \u0000 \u0000This paper was accepted by J. Miguel Villas-Boas, marketing.","PeriodicalId":165362,"journal":{"name":"ERN: Discrete Regression & Qualitative Choice Models (Single) (Topic)","volume":"45 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2014-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127195266","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
We analyze a client's choice of contract in auctions where Dutch law firms compete for routine cases. The distinguishing feature here is that lawyers may submit bids with any fee arrangement they prefer: an hourly rate, a fixed fee or a mixed fee, which is a time-capped fixed fee plus an hourly rate for any additional hours should the case take longer than expected. Furthermore, this format of selling legal services is unusual in that it both forces lawyers to compete directly against each other and allows clients to easily compare these different offers. We empirically estimate a choice model for clients and find robust evidence that hourly rate bids are a client's least-preferred choice. Our findings tentatively contradict lawyers' often made argument that hourly rates are in a client's best interest.
{"title":"What Clients Want: Choices between Lawyers' Offerings","authors":"F. Felsö, S. Onderstal, J. Seldeslachts","doi":"10.2139/ssrn.2591304","DOIUrl":"https://doi.org/10.2139/ssrn.2591304","url":null,"abstract":"We analyze a client's choice of contract in auctions where Dutch law firms compete for routine cases. The distinguishing feature here is that lawyers may submit bids with any fee arrangement they prefer: an hourly rate, a fixed fee or a mixed fee, which is a time-capped fixed fee plus an hourly rate for any additional hours should the case take longer than expected. Furthermore, this format of selling legal services is unusual in that it both forces lawyers to compete directly against each other and allows clients to easily compare these different offers. We empirically estimate a choice model for clients and find robust evidence that hourly rate bids are a client's least-preferred choice. Our findings tentatively contradict lawyers' often made argument that hourly rates are in a client's best interest.","PeriodicalId":165362,"journal":{"name":"ERN: Discrete Regression & Qualitative Choice Models (Single) (Topic)","volume":"3 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2014-02-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126154167","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Dagsvik and Karlstrom (2005) have demonstrated how one can compute Compensating Variation and Compensated Choice Probabilities by means of analytic formulas in the context of discrete choice models. In this paper we offer a new and simplified derivation of the compensated probabilities. Subsequently, we discuss the application of this methodology to compute compensated labor supply responses (elasticities) in a particular discrete choice labor supply model. Whereas the Slutsky equation holds in the case of the standard microeconomic model with deterministic preferences, this is not so in the case of random utility models. When the non-labor income elasticity is negative the Slutsky equation implies that the compensated wage elasticity is higher than the uncompensated one. In contrast, in our random utility model we show empirically that in a majority of cases the uncompensated wage elasticity is in fact the highest one. We also show that when only the deterministic part of the utility function is employed to yield optimal hours and related elasticities, these elasticities are numerically much higher and decline more sharply across deciles than the random utility ones.
{"title":"Compensated Discrete Choice with Particular Reference to Labor Supply","authors":"J. Dagsvik, S. Strøm, Marilena Locatelli","doi":"10.2139/ssrn.2393896","DOIUrl":"https://doi.org/10.2139/ssrn.2393896","url":null,"abstract":"Dagsvik and Karlstrom (2005) have demonstrated how one can compute Compensating Variation and Compensated Choice Probabilities by means of analytic formulas in the context of discrete choice models. In this paper we offer a new and simplified derivation of the compensated probabilities. Subsequently, we discuss the application of this methodology to compute compensated labor supply responses (elasticities) in a particular discrete choice labor supply model. Whereas the Slutsky equation holds in the case of the standard microeconomic model with deterministic preferences, this is not so in the case of random utility models. When the non-labor income elasticity is negative the Slutsky equation implies that the compensated wage elasticity is higher than the uncompensated one. In contrast, in our random utility model we show empirically that in a majority of cases the uncompensated wage elasticity is in fact the highest one. We also show that when only the deterministic part of the utility function is employed to yield optimal hours and related elasticities, these elasticities are numerically much higher and decline more sharply across deciles than the random utility ones.","PeriodicalId":165362,"journal":{"name":"ERN: Discrete Regression & Qualitative Choice Models (Single) (Topic)","volume":"261 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2014-01-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134203984","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
The problem of unstable coecients in the rank-ordered logit model has been traditionally interpreted as a sign that survey respondents fail to provide reliable ranking responses. This paper shows that the problem may embody the inherent sensitivity of the model to stochastic misspecification instead. Even a minor departure from the postulated random utility function can induce the problem, for instance when rank-ordered logit is estimated whereas the true additive disturbance is iid normal over alternatives. Related implications for substantive analyses and further modelling are explored. In general, a well-speci ed random coecient rank-ordered logit model can mitigate, though not eliminate, the problem and produce analytically useful results. The model can also be generalised to be more suitable for forecasting purposes, by accommodating that stochastic misspecification matters less for individuals with more deterministic preferences. An empirical analysis using an Australian nursing job preferences survey shows that the estimates behave in accordance with these implications.
{"title":"The Perceived Unreliability of Rank-Ordered Data: An Econometric Origin and Implications","authors":"H. I. Yoo","doi":"10.2139/ssrn.2172145","DOIUrl":"https://doi.org/10.2139/ssrn.2172145","url":null,"abstract":"The problem of unstable coecients in the rank-ordered logit model has been traditionally interpreted as a sign that survey respondents fail to provide reliable ranking responses. This paper shows that the problem may embody the inherent sensitivity of the model to stochastic misspecification instead. Even a minor departure from the postulated random utility function can induce the problem, for instance when rank-ordered logit is estimated whereas the true additive disturbance is iid normal over alternatives. Related implications for substantive analyses and further modelling are explored. In general, a well-speci ed random coecient rank-ordered logit model can mitigate, though not eliminate, the problem and produce analytically useful results. The model can also be generalised to be more suitable for forecasting purposes, by accommodating that stochastic misspecification matters less for individuals with more deterministic preferences. An empirical analysis using an Australian nursing job preferences survey shows that the estimates behave in accordance with these implications.","PeriodicalId":165362,"journal":{"name":"ERN: Discrete Regression & Qualitative Choice Models (Single) (Topic)","volume":"14 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2012-11-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128853335","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
A direct utility approach can handle multiple discrete/continuous choice outcomes. However, there is a trade-off between allowing correlations between unobserved perceived qualities of two alternatives and computational burden. If we allow the correlations, then we have to do numerical integrations for getting likelihoods, whereas if we give up the correlations, then we can get closed form likelihood. Thus, we suggest a new copula-based direct utility approach that not only allows the correlations but also provides closed form of likelihood. Empirically, we find the existence of the correlations between unobserved qualities of two alternatives. Ignoring the correlations may cause the misunderstanding of the joint-purchases.
{"title":"A Copula-Based Direct Utility Approach with Various Correlations","authors":"D. Jun, Chul Kim","doi":"10.2139/ssrn.2174789","DOIUrl":"https://doi.org/10.2139/ssrn.2174789","url":null,"abstract":"A direct utility approach can handle multiple discrete/continuous choice outcomes. However, there is a trade-off between allowing correlations between unobserved perceived qualities of two alternatives and computational burden. If we allow the correlations, then we have to do numerical integrations for getting likelihoods, whereas if we give up the correlations, then we can get closed form likelihood. Thus, we suggest a new copula-based direct utility approach that not only allows the correlations but also provides closed form of likelihood. Empirically, we find the existence of the correlations between unobserved qualities of two alternatives. Ignoring the correlations may cause the misunderstanding of the joint-purchases.","PeriodicalId":165362,"journal":{"name":"ERN: Discrete Regression & Qualitative Choice Models (Single) (Topic)","volume":"9 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2012-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131295587","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
The mixed logit is a framework for incorporating unobserved heterogeneity in discrete choice models in a general way. These models are difficult to estimate because they result in a complicated incomplete data likelihood. This paper proposes a new approach for estimating mixed logit models. The estimator is easily implemented as iteratively re-weighted least squares: the well known solution for complete data likelihood logits. The main benefit of this approach is that it requires drastically fewer evaluations of the simulated likelihood function, making it significantly faster than conventional methods that rely on numerically approximating the gradient. The method is rooted in a generalized expectation and maximization (GEM) algorithm, so it is asymptotically consistent, efficient, and globally convergent.
{"title":"A Tractable Estimator for General Mixed Multinomial Logit Models","authors":"J. James","doi":"10.26509/WP-201219","DOIUrl":"https://doi.org/10.26509/WP-201219","url":null,"abstract":"The mixed logit is a framework for incorporating unobserved heterogeneity in discrete choice models in a general way. These models are difficult to estimate because they result in a complicated incomplete data likelihood. This paper proposes a new approach for estimating mixed logit models. The estimator is easily implemented as iteratively re-weighted least squares: the well known solution for complete data likelihood logits. The main benefit of this approach is that it requires drastically fewer evaluations of the simulated likelihood function, making it significantly faster than conventional methods that rely on numerically approximating the gradient. The method is rooted in a generalized expectation and maximization (GEM) algorithm, so it is asymptotically consistent, efficient, and globally convergent.","PeriodicalId":165362,"journal":{"name":"ERN: Discrete Regression & Qualitative Choice Models (Single) (Topic)","volume":"85 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2012-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132366957","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
We show that accounting information releases generate large and immediate price impacts, i.e. jumps, in credit default swap (CDS) spreads. Our approach is multivariate, which allows for identification of information events under the presence of confounding news, such as credit events and other simultaneous news arrivals. The economic impact of accounting news releases is twice as large as the impact of credit-related news. Good and bad news impact jumps in CDS spreads asymmetrically, and unscheduled announcements are more likely to cause jumps than scheduled ones. The arrival of accounting information is quickly absorbed in CDS spreads, suggesting efficient price discovery in the CDS market.
{"title":"Accounting Information Releases and CDS Spreads","authors":"Redouane Elkamhi, Kris Jacobs, Hugues Langlois, Chayawat Ornthanalai","doi":"10.2139/ssrn.1874127","DOIUrl":"https://doi.org/10.2139/ssrn.1874127","url":null,"abstract":"We show that accounting information releases generate large and immediate price impacts, i.e. jumps, in credit default swap (CDS) spreads. Our approach is multivariate, which allows for identification of information events under the presence of confounding news, such as credit events and other simultaneous news arrivals. The economic impact of accounting news releases is twice as large as the impact of credit-related news. Good and bad news impact jumps in CDS spreads asymmetrically, and unscheduled announcements are more likely to cause jumps than scheduled ones. The arrival of accounting information is quickly absorbed in CDS spreads, suggesting efficient price discovery in the CDS market.","PeriodicalId":165362,"journal":{"name":"ERN: Discrete Regression & Qualitative Choice Models (Single) (Topic)","volume":"5 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2012-01-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"117154909","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
This paper outlines an approach to improving credit score modeling using random forests and compares random forests with logistic regression. It is shown that on data sets where variables have multicollinearity and complex interrelationships random forests provide a more scientific approach to analyzing variable importance and achieving optimal predictive accuracy. In addition it is shown that random forests should be used in econometric and credit risk models as they provide a powerful too to assess meaning of variables not available in standard regression models and thus allow for more robust findings.
{"title":"Improving the Art, Craft and Science of Economic Credit Risk Scorecards Using Random Forests: Why Credit Scorers and Economists Should Use Random Forests","authors":"Dhruv Sharma","doi":"10.2139/ssrn.1861535","DOIUrl":"https://doi.org/10.2139/ssrn.1861535","url":null,"abstract":"This paper outlines an approach to improving credit score modeling using random forests and compares random forests with logistic regression. It is shown that on data sets where variables have multicollinearity and complex interrelationships random forests provide a more scientific approach to analyzing variable importance and achieving optimal predictive accuracy. In addition it is shown that random forests should be used in econometric and credit risk models as they provide a powerful too to assess meaning of variables not available in standard regression models and thus allow for more robust findings.","PeriodicalId":165362,"journal":{"name":"ERN: Discrete Regression & Qualitative Choice Models (Single) (Topic)","volume":"48 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2011-06-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125569509","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
This paper uses synthetic datasets to classify the conditions in which random forest may outperform more traditional techniques such as logistic regression. We explore the theoretical implications of these experimental findings, and work towards building a theory based approach to data mining. During the course of these experiments we take the simulations where random forests dominate and add additional dimensionality to the data and run logistic regression using the additional attributes through the I* interaction miner algorithm outlined in Sharma 2011. Using the I* procedure with adequate amount of interaction terms the logistic regression can be made to match performance of random forests in the synthetic data sets where random forests dominate (Sharma, 2011). This makes it seem the interaction miner algorithm along with some minimal sufficient amount of interaction and transformations allow logistic regression to match ensemble performance. This implies that, without a certain amount of dimensionality in the data interaction, miner and logistic regression do not benefit from the interactions. Breiman and other work shows Random Forests thrive on dimensionality that said from experiences with various data sets adding additional artificial dimensionality doesn’t help forest (Breiman, 2001). There appears to be some minimum or necessary and sufficient amount of dimensionality after which more information cannot be extracted from the data. The good news is dimensionality can be created using the icreater function which add Tukey’s re-expressions automatically to the data (log, negative reciprocal, and sqrt).
{"title":"Some Experiments Comparing Logistic Regression and Random Forests Using Synthetic Data and the Interaction Miner Algorithm","authors":"Dhruv Sharma","doi":"10.2139/ssrn.1858424","DOIUrl":"https://doi.org/10.2139/ssrn.1858424","url":null,"abstract":"This paper uses synthetic datasets to classify the conditions in which random forest may outperform more traditional techniques such as logistic regression. We explore the theoretical implications of these experimental findings, and work towards building a theory based approach to data mining. During the course of these experiments we take the simulations where random forests dominate and add additional dimensionality to the data and run logistic regression using the additional attributes through the I* interaction miner algorithm outlined in Sharma 2011. Using the I* procedure with adequate amount of interaction terms the logistic regression can be made to match performance of random forests in the synthetic data sets where random forests dominate (Sharma, 2011). This makes it seem the interaction miner algorithm along with some minimal sufficient amount of interaction and transformations allow logistic regression to match ensemble performance. This implies that, without a certain amount of dimensionality in the data interaction, miner and logistic regression do not benefit from the interactions. Breiman and other work shows Random Forests thrive on dimensionality that said from experiences with various data sets adding additional artificial dimensionality doesn’t help forest (Breiman, 2001). There appears to be some minimum or necessary and sufficient amount of dimensionality after which more information cannot be extracted from the data. The good news is dimensionality can be created using the icreater function which add Tukey’s re-expressions automatically to the data (log, negative reciprocal, and sqrt).","PeriodicalId":165362,"journal":{"name":"ERN: Discrete Regression & Qualitative Choice Models (Single) (Topic)","volume":"92 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2011-06-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134266280","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：