High Level Of Prediction Accuracy Research Articles

MODELING A SPATIO-TEMPORAL INDIVIDUAL TRAVEL BEHAVIOR USING GEOTAGGED SOCIAL NETWORK DATA: A CASE STUDY OF GREATER CINCINNATI

Abstract. Understanding individual travel behavior is vital in travel demand management as well as in urban and transportation planning. New data sources including mobile phone data and location-based social media (LBSM) data allow us to understand mobility behavior on an unprecedented level of details. Recent studies of trip purpose prediction tend to use machine learning (ML) methods, since they generally produce high levels of predictive accuracy. Few studies used LSBM as a large data source to extend its potential in predicting individual travel destination using ML techniques. In the presented research, we created a spatio-temporal probabilistic model based on an ensemble ML framework named “Random Forests” utilizing the travel extracted from geotagged Tweets in 419 census tracts of Greater Cincinnati area for predicting the tract ID of an individual’s travel destination at any time using the information of its origin. We evaluated the model accuracy using the travels extracted from the Tweets themselves as well as the travels from household travel survey. The Tweets and survey based travels that start from same tract in the south western parts of the study area is more likely to select same destination compare to the other parts. Also, both Tweets and survey based travels were affected by the attraction points in the downtown of Cincinnati and the tracts in the north eastern part of the area. Finally, both evaluations show that the model predictions are acceptable, but it cannot predict destination using inputs from other data sources as precise as the Tweets based data.

Emotional eating and cognitive conflicts as predictors of binge eating disorder in patients with obesity

Background/Objectives: Binge Eating Disorder (BED) is often associated with obesity. In order to identify the variables that allow to better detect the presence of BED, people with overnutrition were compared with and without BED in the presence of cognitive conflicts, eating symptoms and anxious-depressive symptoms. The inclusion of cognitive conflicts had been relevant in bulimia studies but had not been investigated with respect to BED. Method: Two groups with obesity were evaluated, one without BED (OB, n=54) and the other with BED (OB-BED, n=48), using a social-demographic questionnaire as well as a semi-structured interview to assess BED, questionnaires (DASS-21, EDE-Q, EEQ) and the Repertory Grid Technique. Results: Overall, the OB-BED group presented more conflicts and more symptoms. The model that best differentiated between the groups included emotional eating and level of cognitive conflicts, correctly classifying 91.4% of the sample. Conclusion: These results highlight the role played by cognitive conflicts and emotional eating as differentiating elements between OB and OB-BED, with a high level of predictive accuracy.

Association and effects of gene-gene interactions between peroxisome proliferator-activated receptor and pulse pressure

Objective: To investigate the association between ten single nucleotide polymorphisms (SNPs) in the peroxisome proliferator-activated receptors and pulse pressure (PP) as well as the relationships between gene-gene interaction between PPARα/δ/γ genes and PP. Methods: A total of 820 subjects, with 550 females and 270 males, were recruited from a cohort study of "Prevention of Metabolic Syndrome and Multi-metabolic Disorders in Jiangsu Province of China Study (PMMJS)" . Ten SNPs of PPARα/δ/γ genes were selected. GMDR software (version 1.0.1) was used to evaluate the gene-gene interactions among PPARs SNPs associated with PP. Results: The mean levels of PP in people with mutant genotype of rs1805192 in PPARγ genes (PA+AA) showed a significant increase by 1.341 mmHg (95%CI: 0.431-2.252 mmHg) when compared to the persons with wild genotype (PP). In the subgroup of subjects with more than 30 mmHg levels of PP, a six-locus model comprised rs135539 of PPARα, rs2016520 of PPARδ, rs10865710, rs1805192, rs709158 and rs3856806 of PPARγ showed a highest level of prediction accuracy (0.577) and displayed a better cross-validation consistency (10/10). In the subgroup of subjects with less than 40 mmHg levels of PP, a two-locus model was statistically associated with PP with 0.628 of prediction accuracy and 10/10 of cross-validation consistency. Conclusion: PPARγ rs1805192 was associated with the occurrence of PP. Gene-gene interactions among rs135539 of PPARα, rs2016520 of PPARδ, rs10865710, rs1805192, rs709158 and rs3856806 of PPARγ were all significantly related to PP.

A sparse partitioned-regression model for nonlinear system–environment interactions

ABSTRACTThis article focuses on the modeling of nonlinear interactions between the design and operational variables of a system and the multivariate outside environment in predicting the system's performance. We propose a Sparse Partitioned-Regression (SPR) model that automatically searches for a partition of the environmental variables and fits a sparse regression within each subdivision of the partition, in order to fulfill an optimal criterion. Two optimal criteria are proposed, a penalized and a held-out criterion. We study the theoretical properties of SPR by deriving oracle inequalities to quantify the risks of the penalized and held-out criteria in both prediction and classification problems. An efficient recursive partition algorithm is developed for model estimation. Extensive simulation experiments are conducted to demonstrate the better performance of SPR compared with competing methods. Finally, we present an application of using building design and operational variables, outdoor environmental variables, and their interactions to predict energy consumption based on the Department of Energy's EnergyPlus data sets. SPR produces a high level of prediction accuracy. The result of the application also provides insights into the design, operation, and management of energy-efficient buildings.

Genomic Selection for Drought Tolerance Using Genome-Wide SNPs in Maize.

Traditional breeding strategies for selecting superior genotypes depending on phenotypic traits have proven to be of limited success, as this direct selection is hindered by low heritability, genetic interactions such as epistasis, environmental-genotype interactions, and polygenic effects. With the advent of new genomic tools, breeders have paved a way for selecting superior breeds. Genomic selection (GS) has emerged as one of the most important approaches for predicting genotype performance. Here, we tested the breeding values of 240 maize subtropical lines phenotyped for drought at different environments using 29,619 cured SNPs. Prediction accuracies of seven genomic selection models (ridge regression, LASSO, elastic net, random forest, reproducing kernel Hilbert space, Bayes A and Bayes B) were tested for their agronomic traits. Though prediction accuracies of Bayes B, Bayes A and RKHS were comparable, Bayes B outperformed the other models by predicting highest Pearson correlation coefficient in all three environments. From Bayes B, a set of the top 1053 significant SNPs with higher marker effects was selected across all datasets to validate the genes and QTLs. Out of these 1053 SNPs, 77 SNPs associated with 10 drought-responsive transcription factors. These transcription factors were associated with different physiological and molecular functions (stomatal closure, root development, hormonal signaling and photosynthesis). Of several models, Bayes B has been shown to have the highest level of prediction accuracy for our data sets. Our experiments also highlighted several SNPs based on their performance and relative importance to drought tolerance. The result of our experiments is important for the selection of superior genotypes and candidate genes for breeding drought-tolerant maize hybrids.

Predictors of performance in a 4-h mountain-bike race

ABSTRACTThis study aimed to cross validate previously developed predictive models of mountain biking performance in a new cohort of mountain bikers during a 4-h event (XC4H). Eight amateur XC4H cyclists completed a multidimensional assessment battery including a power profile assessment that consisted of maximal efforts between 6 and 600 s, maximal hand grip strength assessments, a video-based decision-making test as well as a XC4H race. A multiple linear regression model was found to predict XC4H performance with good accuracy (R2 = 0.99; P < 0.01). This model consisted of relative to total cycling mass (body mass including competition clothing and bicycle mass), maximum power output sustained over 60 s relative to total cycling mass, peak left hand grip strength and two-line decision-making score. Previous models for Olympic distance MTB performance demonstrated merit (R2 = 0.93; P > 0.05) although subtle changes improved the fit, significance and normal distribution of residuals within the model (R2 = 0.99; P < 0.01), highlighting differences between the disciplines. The high level of predictive accuracy of the new XC4H model further supports the use of a multidimensional approach in predicting MTB performance. The difference between the new, XC4H and previous Olympic MTB predictive models demonstrates subtle differences in physiological requirements and performance predictors between the two MTB disciplines.

Mathematical modeling of ovine footrot in the UK: the effect of Dichelobacter nodosus and Fusobacterium necrophorum on the disease dynamics

Dichelobacter nodosus is a virulent, invasive, anaerobic bacterium that is believed to be the causative agent of ovine footrot, an infectious bacterial disease of sheep that causes lameness. Another anaerobe, Fusobacterium necrophorum, has been intimately linked with the disease occurrence and severity. Here we examine data from a longitudinal study of footrot on one UK farm, including quantitative PCR (qPCR) estimates of bacterial load of D. nodosus and F. necrophorum. The data is at foot level; all feet were monitored for five weeks assessing disease severity (healthy, interdigital dermatitis (ID), or severe footrot (SFR)) and bacterial load (number of bacteria/swab). We investigate the role of D. nodosus and F. necrophorum in the progress of the disease using a continuous-time Markov model with 12 different states characterising the foot. The transition rates between the adjacent states are the (34) model parameters, these are determined using Metropolis Hasting MCMC. Our aim is to determine the predictive relationship between past and future D. nodosus and F. necrophorum load and disease states. We demonstrate a high level of predictive accuracy at the population level for the D. nodosus model, although the dynamics of individual feet is highly stochastic. However, we note that this predictive accuracy at population level is only high in more diseased states for F. necrophorum model. This supports our hypothesis that D. nodosus load and status of the foot work in combination to give rise to severe footrot and lameness, and that D. nodosus load plays the primary role in the initiation and progression of footrot, while F. necrophorum load rather increases disease severity of SFR.

Assessment of food compositional parameters by means of a Waveguide Vector Spectrometer

A Waveguide Vector Spectrometer (WVS), operating in the frequency range 1.6–2.7 GHz, was designed, set up and tested for rapid assessment of main chemical-physical properties of a plurality of food products. The system, completely integrated, includes a waveguide, a control unit, a signal generator, a gain/phase comparator, a power supply, and a USB port (for control and data transfer). The information contained in the gain and phase waveforms related to the product dielectric properties are exploited for the prediction of different compounds characterising both simple liquid solutions and more complex food products.By processing spectral data with Partial Least Square Regression (PLS) algorithm, high levels of prediction accuracy were observed for all the tested products (R2 values from 0.940 to 0.999, in validation). Generally, both gain and phase waveforms appeared to be effective, for the selected foods, in terms of prediction accuracy. On the whole, the proposed system seems able to assess the content of varied substances both on simple and complex matrices. Simple prospective changes of the sample holder make the equipment potentially suitable for on-line monitoring of the quality of food products.

A machine learning approach to geochemical mapping

Geochemical maps provide invaluable evidence to guide decisions on issues of mineral exploration, agriculture, and environmental health. However, the high cost of chemical analysis means that the ground sampling density will always be limited. Traditionally, geochemical maps have been produced through the interpolation of measured element concentrations between sample sites using models based on the spatial autocorrelation of data (e.g. semivariogram models for ordinary kriging). In their simplest form such models fail to consider potentially useful auxiliary information about the region and the accuracy of the maps may suffer as a result. In contrast, this study uses quantile regression forests (an elaboration of random forest) to investigate the potential of high resolution auxiliary information alone to support the generation of accurate and interpretable geochemical maps. This paper presents a summary of the performance of quantile regression forests in predicting element concentrations, loss on ignition and pH in the soils of south west England using high resolution remote sensing and geophysical survey data.Through stratified 10-fold cross validation we find the accuracy of quantile regression forests in predicting soil geochemistry in south west England to be a general improvement over that offered by ordinary kriging. Concentrations of immobile elements whose distributions are most tightly controlled by bedrock lithology are predicted with the greatest accuracy (e.g. Al with a cross-validated R2 of 0.79), while concentrations of more mobile elements prove harder to predict. In addition to providing a high level of prediction accuracy, models built on high resolution auxiliary variables allow for informative, process based, interpretations to be made. In conclusion, this study has highlighted the ability to map and understand the surface environment with greater accuracy and detail than previously possible by combining information from multiple datasets. As the quality and coverage of remote sensing and geophysical surveys continue to improve, machine learning methods will provide a means to interpret the otherwise-uninterpretable.

Quasi steady-state aerodynamic model development for race vehicle simulations

Presented in this paper is a procedure to develop a high fidelity quasi steady-state aerodynamic model for use in race car vehicle dynamic simulations. Developed to fit quasi steady-state wind tunnel data, the aerodynamic model is regressed against three independent variables: front ground clearance, rear ride height, and yaw angle. An initial dual range model is presented and then further refined to reduce the model complexity while maintaining a high level of predictive accuracy. The model complexity reduction decreases the required amount of wind tunnel data thereby reducing wind tunnel testing time and cost. The quasi steady-state aerodynamic model for the pitch moment degree of freedom is systematically developed in this paper. This same procedure can be extended to the other five aerodynamic degrees of freedom to develop a complete six degree of freedom quasi steady-state aerodynamic model for any vehicle.

Brain Biomarkers of Vulnerability and Progression to Psychosis.

Identifying predictors and elucidating the fundamental mechanisms underlying onset of psychosis are critical for the development of targeted preemptive interventions. This article presents a selective review of findings on risk prediction algorithms and potential mechanisms of onset in youth at clinical high-risk for psychosis, focusing principally on recent findings of the North American Prodrome Longitudinal Study (NAPLS). Multivariate models incorporating risk factors from clinical, demographic, neurocognitive, and psychosocial assessments achieve high levels of predictive accuracy when applied to individuals who meet criteria for a prodromal risk syndrome. An individualized risk calculator is available to scale the risk for newly ascertained cases, which could aid in clinical decision making. At risk individuals who convert to psychosis show elevated levels of proinflammatory cytokines, as well as disrupted resting state thalamo-cortical functional connectivity at baseline, compared with those who do not. Further, converters show a steeper rate of gray matter reduction, most prominent in prefrontal cortex, that in turn is predicted by higher levels of inflammatory markers at baseline. Microglia, resident immune cells in the brain, have recently been discovered to influence synaptic plasticity in health and impair plasticity in disease. Processes that modulate microglial activation may represent convergent mechanisms that influence brain dysconnectivity and risk for onset of psychosis and thus may be targetable in developing and testing preventive interventions.

An attempt of a crystal ball for hotels in a gaming destination, Macao

Purpose – The purpose of this paper is to consider the issue of forecasting hotel room rate with data from 2004 onwards and the forecast horizons of 91 observations. Design/methodology/approach – This study employs a set of time series data (91 observations) on average monthly hotel room rates to generate an Autoregressive Integrated Moving Average Models (ARIMA) forecasting model. Findings – Through the employment of 74 percent observations, with 26 percent withhold for evaluation checking, six best models are identified from 50 models under study. The final model reports a high level of predictive accuracy and provides useful insights for hoteliers to form business strategies. Originality/value – This research provides a well-defined model to forecast the room rate of three-star hotels in the city. The research findings provide good ground for strategic management of the industry, in which there is an imbalance between demand and supply of hotel accommodations. In addition, being the first of its kind in one of the largest gaming revenue generation city in the world, this research provides valuable information for further research of its kind in the future.

Can we improve the targeting of respiratory syncytial virus (RSV) prophylaxis in infants born 32–35 weeks’ gestational age with more informed use of risk factors?

Objective: To evaluate the key risk factors for respiratory syncytial virus (RSV) hospitalisation in 32–35 weeks’ gestational age (wGA) infants.Methods: Published risk factors were assessed for predictive accuracy (area under the receiver operating characteristic curve [ROC AUC]) and for number needed to treat (NNT).Results: Key risk factors included: proximity of birth to the RSV season; having siblings; crowding at home; day care; smoking; breast feeding; small for GA; male gender; and familial wheezing/eczema. Proximity of birth to the RSV season appeared the most predictive. Risk factors models from Europe and Canada were found to have a high level of predictive accuracy (ROC AUC both >0.75; NNT for European model 9.5). A model optimised for three risk factors (birth ±10 weeks from start of RSV season, number of siblings ≥2 years and breast feeding for ≤2 months) had a similar level of prediction (ROC AUC: 0.776; NNT: 10.2). An example two-risk factor model (day care attendance and living with ≥2 siblings <5 years old) had a lower level of predictive accuracy (ROC AUC: 0.55; NNT: 26).Conclusions: An optimised combination of risk factors has the potential to improve the identification of 32–35 wGA infants at heightened risk of RSV hospitalisation.

A comparative investigation of modern feature selection and classification approaches for the analysis of mass spectrometry data

Many analytical approaches such as mass spectrometry generate large amounts of data (input variables) per sample analysed, and not all of these variables are important or related to the target output of interest. The selection of a smaller number of variables prior to sample classification is a widespread task in many research studies, where attempts are made to seek the lowest possible set of variables that are still able to achieve a high level of prediction accuracy; in other words, there is a need to generate the most parsimonious solution when the number of input variables is huge but the number of samples/objects are smaller. Here, we compare several different variable selection approaches in order to ascertain which of these are ideally suited to achieve this goal. All variable selection approaches were applied to the analysis of a common set of metabolomics data generated by Curie-point pyrolysis mass spectrometry (Py-MS), where the goal of the study was to classify the Gram-positive bacteria Bacillus. These approaches include stepwise forward variable selection, used for linear discriminant analysis (LDA); variable importance for projection (VIP) coefficient, employed in partial least squares-discriminant analysis (PLS-DA); support vector machines-recursive feature elimination (SVM-RFE); as well as the mean decrease in accuracy and mean decrease in Gini, provided by random forests (RF). Finally, a double cross-validation procedure was applied to minimize the consequence of overfitting. The results revealed that RF with its variable selection techniques and SVM combined with SVM-RFE as a variable selection method, displayed the best results in comparison to other approaches.

ANALISIS VARIABEL KANONIK BIPLOT UNTUK BANK UMUM DI JAWA TENGAH

Bank Competition in Indonesia increase due to good economic growth and the improvement of the social middle class in Indonesia. Increased bank raises the fierce competition between banks and internal banks themselves. This makes the management of the bank should work seriously to maintain its existence. In this case the assessment of the bank become very important in the banking business to survive in today's banking industry. This study was conducted to determine the competitive commercial banks operating in Central Java with the Canonical Variate Analysis (CVA) Biplot. This analysis can be applied to find out information about the relative position, the similarity between the object characteristics and diversity of variables in the three groups of commercial banks in Central Java, namely state-owned banks, private banks and private banks Non Foreign Exchange, based on the health aspects of the bank. The results obtained are the banks in each group had different characteristics shown in the relative position of the already well-separated in the resulting biplot. Variables that tend to influence the grouping of commercial banks are Capital Adequacy Ratio (CAR). The total assets is variable with the highest level of prediction accuracy on each bank. Keywords : Health Aspects of the Bank , Commercial Banks, Canonical Variate Analysis (C V A ) Biplo t .

Validation of and revision to the VRAG and SORAG: The Violence Risk Appraisal Guide—Revised (VRAG-R).

The violence risk appraisal guide (VRAG) was developed in the early 1990s, and approximately 60 replications around the world have shown its utility for the appraisal of violence risk among correctional and psychiatric populations. At the same time, authorities (e.g., Dawes, Faust, & Meehl, 1989) have argued that tools should be periodically evaluated to see if they need to be revised. In the present study, we evaluated the accuracy of the VRAG in a sample of 1,261 offenders, fewer than half of whom were participants in the development sample, then developed and validated a revised and easier-to-score instrument (the VRAG-R). We examined the accuracy of both instruments over fixed durations of opportunity ranging from 6 months to 49 years and examined outcome measures pertaining to the overall number, severity, and imminence of violent recidivism. Both instruments were found to predict dichotomous violent recidivism overall and at various fixed follow-ups with high levels of predictive accuracy (receiver operating characteristic areas of approximately .75) and to significantly predict other violent outcomes.

Evaluating ensemble forecasts of plant species distributions under climate change

Species distributions models (SDMs) are commonly used to assess potential species’ range shifts or extinction risk under climate change. It has been suggested that the use of ensemble forecasts, where a variety of model algorithms are used to generate consensus predictions, are preferred to individual SDMs by avoiding bias or prediction error inherent in a single modeling approach. Whereas several studies have assessed the performance of ensemble predictions using cross-validation or data-partitioning approaches, few studies have assessed the predictive accuracy of ensemble forecasts under climate change by using temporally independent model validation data. We used five SDM approaches to develop consensus forecasts of distributions of 145 vascular plant species from California in the 1930s and tested their projections against current distributions, a span of approximately 75 years. When evaluated with a portion of the model training data, consensus forecasts were highly accurate with an average AUC value of 0.97. False positive and false negative error rates were also low, exhibiting similar performance to random forest models. However, when evaluated with temporally independent data, the accuracy of consensus forecasts was similar to that of generalized linear and generalized additive models, with an average AUC value of 0.83. Our results suggest that the high levels of predictive accuracy exhibited by consensus forecasts when using data partitioning approaches may not reflect their performance when predicting temporally independent data. We contend that consensus forecasts may not represent the best approach for predicting species distributions under future climatic change, as they may not provide superior predictive accuracy in novel temporal domains compared to traditional modeling approaches that more readily lend themselves to ecological interpretation of model structure.

Forest mapping and monitoring using active 3D remote sensing

The main aim in forest mapping and monitoring is to produce accurate information for forest managers with the use of efficient methodologies. For example, it is important to locate harvesting sites and stands where forest operations should be carried out as well as to provide updates regarding forest growth, among other changes in forest structure. In recent years, remote sensing (RS) has taken a significant technological leap forward. It has become possible to acquire three-dimensional (3D), spatially accurate information from forest resources using active RS methods. In practical applications, mainly 3D information produced by airborne laser scanning (ALS) has opened up groundbreaking potential in natural resource mapping and monitoring. In addition to ALS, new satellite radars are also capable of acquiring spatially accurate 3D information. The main objectives of the present study were to develop 3D RS methodologies for large-area forest mapping and monitoring applications. In substudy I, we aim to map harvesting sites, while in substudy II, we monitor changes in the forest canopy structure. In studies III-V, efficient mapping and monitoring applications were developed and tested. In substudy I, we predicted plot-level thinning maturity within the next 10-year planning period. Stands requiring immediate thinning were located with an overall accuracy of 83%-86% depending on the prediction method applied. The respective prediction accuracy for stands reaching thinning maturity within the next 10 years was 70%-79%. Substudy II addressed natural disturbance monitoring that could be linked to forest management planning when an ALS time series is available. The accuracy of the damaged canopy cover area estimate varied between -16.4% to 5.4%. Substudy II showed that changes in the forest canopy structure can be monitored with a rather straightforward method by contrasting bi-temporal canopy height models. In substudy III, we developed a RS-based forest inventory method where single-tree RS is used to acquire modelling data needed in area-based predictions. The method uses ALS data and is capable of producing accurate stand variable estimates even at the sub-compartment level. The developed method could be applied in areas with sparse road networks or when the costs of fieldwork must be minimized. The method is especially suitable for largearea biomass or stem volume mapping. Based on substudy IV, the use of stereo synthetic aperture radar (SAR) satellite data in the prediction of plotlevel forest variables appears to be promising for large-area applications. In the best case, the plot-level stem volume (VOL) was predicted with a relative error (RMSE%) of 34.9%. Typically, such a high level of prediction accuracy cannot be obtained using spaceborne RS data. Then, in substudy V, we compared the aboveground biomass and VOL estimates derived by radargrammetry to the ALS estimates. The difference between the estimation accuracy of ALS–based and TerraSAR X–based features was smaller than in any previous study in which ALS and different kinds of SAR materials have been compared. In this thesis, forest mapping and monitoring applications using active 3D RS were developed. Spatially accurate 3D RS enables the mapping of harvesting sites, the monitoring of changes in the canopy structure and even the making of a fully RS-based forest inventory. ALS is carried out at relatively low altitudes, which makes it relatively expensive per area unit, and other RS materials are still needed. Spaceborne stereo radargrammetry proved to be a promising technique to acquire additional 3D RS data efficiently as long as an accurate digital terrain model is available as a ground-surface reference.

A simple fatigue performance model of asphalt mixtures based on fracture energy

A simple fatigue prediction model of asphalt mixtures based on fracture energy was proposed in this study. The fracture energy density obtained from indirect tension (IDT) test was employed in the fatigue model to consider the resisting energy of the asphalt mixtures against fracture damage. Four point bending beam fatigue tests were performed at various strain levels to characterize the fatigue performance of the two different asphalt mixtures with varying binder contents. Using the fatigue tests data, various types of fatigue models were evaluated. Traditional strain or dissipated energy based fatigue models are simple but strongly dependent on the material types and unable to identify fatigue performance for various asphalt mixtures. A fatigue model based on the ratio of dissipated energy change is independent of the material types and shows relatively high levels of prediction accuracy. However, this fatigue model needs cyclic fatigue tests to characterize the model coefficients. Unlike the fatigue models mentioned above, the newly proposed fatigue model is independent of binder types, binder contents, strain levels and volumetric properties. The proposed fatigue model is able to reasonably predict the fatigue lives of various asphalt mixtures without changing the model coefficients. Moreover, the material properties used in the model can be obtained from the simple IDT test.

Using partial least squares and support vector machines for bankruptcy prediction

The evaluation of corporate financial distress has attracted significant global attention as a result of the increasing number of worldwide corporate failures. There is an immediate and compelling need for more effective financial distress prediction models. This paper presents a novel method to predict bankruptcy. The proposed method combines the partial least squares (PLS) based feature selection with support vector machine (SVM) for information fusion. PLS can successfully identify the complex nonlinearity and correlations among the financial indicators. The experimental results demonstrate its superior predictive ability. On the one hand, the proposed model can select the most relevant financial indicators to predict bankruptcy and at the same time identify the role of each variable in the prediction process. On the other hand, the proposed model’s high levels of prediction accuracy can translate into benefits to financial organizations through such activities as credit approval, and loan portfolio and security management.