Stable Prediction Model Research Articles

With string model to time series forecasting

Overwhelming majority of econometric models applied on a long term basis in the financial forex market do not work sufficiently well. The reason is that transaction costs and arbitrage opportunity are not included, as this does not simulate the real financial markets. Analyses are not conducted on the non equidistant date but rather on the aggregate date, which is also not a real financial case. In this paper, we would like to show a new way how to analyze and, moreover, forecast financial market. We utilize the projections of the real exchange rate dynamics onto the string-like topology in the OANDA market. The latter approach allows us to build the stable prediction models in trading in the financial forex market. The real application of the multi-string structures is provided to demonstrate our ideas for the solution of the problem of the robust portfolio selection. The comparison with the trend following strategies was performed, the stability of the algorithm on the transaction costs for long trade periods was confirmed.

Analysis of Maize (Zea mays) Kernel Density and Volume Using Microcomputed Tomography and Single-Kernel Near-Infrared Spectroscopy

Maize kernel density affects milling quality of the grain. Kernel density of bulk samples can be predicted by near-infrared reflectance (NIR) spectroscopy, but no accurate method to measure individual kernel density has been reported. This study demonstrates that individual kernel density and volume are accurately measured using X-ray microcomputed tomography (μCT). Kernel density was significantly correlated with kernel volume, air space within the kernel, and protein content. Embryo density and volume did not influence overall kernel density. Partial least-squares (PLS) regression of μCT traits with single-kernel NIR spectra gave stable predictive models for kernel density (R(2) = 0.78, SEP = 0.034 g/cm(3)) and volume (R(2) = 0.86, SEP = 2.88 cm(3)). Density and volume predictions were accurate for data collected over 10 months based on kernel weights calculated from predicted density and volume (R(2) = 0.83, SEP = 24.78 mg). Kernel density was significantly correlated with bulk test weight (r = 0.80), suggesting that selection of dense kernels can translate to improved agronomic performance.

Near Infrared Spectral Linearisation in Quantifying Soluble Solids Content of Intact Carambola

This study presents a novel application of near infrared (NIR) spectral linearisation for measuring the soluble solids content (SSC) of carambola fruits. NIR spectra were measured using reflectance and interactance methods. In this study, only the interactance measurement technique successfully generated a reliable measurement result with a coefficient of determination of (R2) = 0.724 and a root mean square error of prediction for (RMSEP) = 0.461° Brix. The results from this technique produced a highly accurate and stable prediction model compared with multiple linear regression techniques.

Stability of Genomic Selection prediction models across ages and environments

Background A tree breeding program is characterized by long generation intervals which, over time, result in a much smaller number of breeding cycles when compared to annual crops. Moreover, most economically important traits in a tree-breeding program are quantitatively inherited, display low heritability and are expressed late in the life cycle. Genomic Selection (GS) is expected to be particularly valuable for tree species, leading to shorter generation intervals and improved genetic gain over time. The main factors that affect the accuracy of GS prediction models are the level of linkage disequilibrium (LD) in the training population, the training population size, the heritability of the trait and the number of QTL regulating its variation. However, it is yet largely unknown how stable prediction models are across environments and different ages. This knowledge is critical for tree breeders that wish to use genomic selection in their genetic improvement program. Here, we report the first assessment of the utility of genomic selection in a conifer species. We developed prediction models for growth traits measured at multiple sites, to evaluate the impact of genotype by environment interactions in their accuracy. Training populations were also measured over multiple ages and models were developed to assess their value in predicting breeding values later in the lifecycle.

Predictive Model of Rat Reproductive Toxicity from ToxCast High Throughput Screening1

The U.S. Environmental Protection Agency's ToxCast research program uses high throughput screening (HTS) for profiling bioactivity and predicting the toxicity of large numbers of chemicals. ToxCast Phase I tested 309 well-characterized chemicals in more than 500 assays for a wide range of molecular targets and cellular responses. Of the 309 environmental chemicals in Phase I, 256 were linked to high-quality rat multigeneration reproductive toxicity studies in the relational Toxicity Reference Database. Reproductive toxicants were defined here as having achieved a reproductive lowest-observed-adverse-effect level of less than 500 mg kg(-1) day(-1). Eight-six chemicals were identified as reproductive toxicants in the rat, and 68 of those had sufficient in vitro bioactivity to model. Each assay was assessed for univariate association with the identified reproductive toxicants. Significantly associated assays were linked to gene sets and used for the subsequent predictive modeling. Using linear discriminant analysis and fivefold cross-validation, a robust and stable predictive model was produced capable of identifying rodent reproductive toxicants with 77% ± 2% and 74% ± 5% (mean ± SEM) training and test cross-validation balanced accuracies, respectively. With a 21-chemical external validation set, the model was 76% accurate, further indicating the model's potential for prioritizing the many thousands of environmental chemicals with little to no hazard information. The biological features of the model include steroidal and nonsteroidal nuclear receptors, cytochrome P450 enzyme inhibition, G protein-coupled receptors, and cell signaling pathway readouts-mechanistic information suggesting additional targeted, integrated testing strategies and potential applications of in vitro HTS to risk assessment.

Do Transcriptomic Markers Provide Significant and Stable Information in Addition to Standard Markers, for the Prediction of Pathological Complete Response in Breast Cancer?

Abstract Background: The aim of this study was to investigate and quantify the contribution of transcriptomic markers, in addition to strong predictors such as oestrogen receptor status, to the prediction of pathological complete response (pCR) in locally advanced breast cancer.Patients: The RNA profiles were analyzed using U133 plus 2.0 Affymetrix. We included 189 patients out of 340 patients entered in a neoadjuvant chemotherapy trial for large operable and locally advanced breast cancer. After four cycles of epirubicin–cyclophosphamide, patients were randomly allocated to four cycles of docetaxel with or without celecoxib for patients with HER2-negative tumors, and docetaxel with or without trastuzumab for patients with HER2-positive tumors, respectively. Proportions of pCR in each group were equal to 0.12, 0.16, 0.15 and 0.24 respectively. Patients who received trastuzumab (N=36) were discard from our example, in order to deal with similar proportions of pCR.Methods: The whole sample was divided into a training set (N=81) and a validation set (N=72). Using the training set, two predictive models were built using multivariate logistic regression models. In the first model (M1), usual clinical and biological significant markers were included. In the second model (M2), in addition to the significant parameters of M1, significant transcriptomic variables were included. Diagnostics of both predictive models were assessed on the validation set through sensitivity and specificity estimates. Simulations were performed to investigate stability of model M2.Results: In M1, oestrogen receptor status and tumor size were found to have a strong predictive role in the prediction of pCR. In addition to these classical markers, genes belonging to biological pathways involved in proliferation and microtubule stabilization appeared to have a strong role in the prediction of pCR (model M2). Validation of M1 on the validation set provided 70% of sensitivity and 86% of specificity. Validation of M2 on the validation set yielded to a better sensitivity of 80% and a specificity of 81%. Using simulations, we showed that several different predictive models M2 yielded to similar performances on the validation set. Conclusion: Our study showed that transcriptomic markers provided significant information in addition to usual biological markers for the prediction of pCR. In addition, predictive model with both usual and transcriptomic markers may lead in an improvement of the classification performances. However, as illustrated by simulations, predictive models with both classical and transcriptomic markers are not exclusive. The contribution of transcriptomic data for the prediction of pCR is straightforward, but finding a stable predictive model remains a great challenge.Supported by PHRC AOM/2OO2/02117, Pfizer inc., Roche, sanofi-aventis.ISRCTN10059974 Citation Information: Cancer Res 2009;69(24 Suppl):Abstract nr 2035.

Membrane-Interaction Quantitative Structure−Activity Relationship (MI-QSAR) Analyses of Skin Penetration Enhancers

Membrane-interaction quantitative structure-activity relationship (MI-QSAR) models for two skin penetration enhancer data sets of 61 and 42 compounds were constructed and compared to QSAR models constructed for the same two data sets using only classic intramolecular QSAR descriptors. These two data sets involve skin penetration enhancement of hydrocortisone and hydrocortisone acetate, and the enhancers are generally similar in structure to lipids and surfactants. A new MI-QSAR descriptor, the difference in the integrated cylindrical distribution functions over the phospholipid monolayer model, in and out of the presence of the skin penetration enhancer, DeltaSigma h(r), was developed. This descriptor is dominant in the optimized MI-QSAR models of both training sets studied and greatly reduces the size and complexity of the MI-QSAR models as compared to those QSAR models developed using the classic intramolecular descriptors. The MI-QSAR models indicate that good penetration enhancers make bigger "holes" in the monolayer and are less aqueous-soluble, so as to preferentially enter the monolayer, than are poor penetration enhancers. The skin penetration enhancer thus alters the structure and organization of the monolayer. This space and time alteration in the structure and dynamics of the membrane monolayer is captured by DeltaSigma h(r) and is simplistically referred to as "holes" in the monolayer. The MI-QSAR models explain 70-80% of the variance in skin penetration enhancement across each of the two training sets and are stable predictive models using accepted diagnostic measures of robustness and predictivity.

Predictors of Depression after Stroke

Although depression is common after stroke, there is uncertainty over its etiology and risk factors, which complicates management. Knowledge of the predictors of depression associated with stroke may allow for the better targeting of therapy, both prevention and treatment. We undertook a systematic review of all published, nonexperimental, population-, hospital-, and rehabilitation-based stroke studies (to June 2004) with prospective, consecutive patient recruitment undertaken to identify variables associated with depressive symptoms (or "illness") after stroke. Assessments were made of the quality of studies including the validity of prognostic models. Data were available from 3 population-based studies including 492 patients, 8 hospital-based studies including 15,272 patients, and 9 rehabilitation-based studies including 2170 patients. Physical disability, stroke severity and cognitive impairment were consistently associated with depression. In addition to the common problem of selection bias, major limitations of these studies included variable selection and poor statistical quality and reporting; small sample sizes meant that only a limited range of variables were analyzed in multivariate models. There is a paucity of well-designed studies of sufficient size to allow stable multivariate predictive models of depression after stroke to be developed. Other than showing that depression is associated with more severe strokes, current evidence does not allow for ready identification of patients most at risk of developing this important complication of stroke.

A new ART-counterpropagation neural network for solving a forecasting problem

This study presents a novel Adaptive resonance theory-Counterpropagation neural network (ART-CPN) for solving forecasting problems. The network is based on the ART concept and the CPN learning algorithm for constructing the neural network. The vigilance parameter is used to automatically generate the nodes of the cluster layer for the CPN learning process. This process improves the initial weight problem and the adaptive nodes of the cluster layer (Kohonen layer). ART-CPN involves real-time learning and is capable of developing a more stable and plastic prediction model of input patterns by self-organization. The advantages of ART-CPN include the ability to cluster, learn and construct the network model for forecasting problems. The network was applied to solve the real forecasting problems. The learning algorithm revealed better learning efficiency and good prediction performance. q 2004 Elsevier Ltd. All rights reserved.

Preoperative prediction of severe postoperative pain

We developed and validated a prediction rule for the occurrence of early postoperative severe pain in surgical inpatients, using predictors that can be easily documented in a preoperative setting. A cohort of surgical inpatients ( n=1416) undergoing various procedures except cardiac surgery and intracranial neurosurgery in a University Hospital were studied. Preoperatively the following predictors were collected: age, gender, type of scheduled surgery, expected incision size, blood pressure, heart rate, Quetelet index, the presence and severity of preoperative pain, health-related quality of life the (SF-36), Spielberger's State–Trait Anxiety Inventory (STAI) and the Amsterdam Preoperative Anxiety and Information Scale (APAIS). The outcome was the presence of severe postoperative pain (defined as Numeric Rating Scale ≥8) within the first hour postoperatively. Multivariate logistic regression in combination with bootstrapping techniques (as a method for internal validation) was used to derive a stable prediction model. Independent predictors of severe postoperative pain were younger age, female gender, level of preoperative pain, incision size and type of surgery. The area under the receiver operator characteristic (ROC) curve was 0.71 (95% CI: 0.68–0.74). Adding APAIS scores (measures of preoperative anxiety and need for information), but not STAI, provided a slightly better model (ROC area 0.73). The reliability of this extended model was good (Hosmer and Lemeshow test p-value 0.78). We have demonstrated that severe postoperative pain early after awakening from general anesthesia can be predicted with a scoring rule, using a small set of variables that can be easily obtained from all patients at the preoperative visit. Before this internally validated preoperative prediction rule can be applied in clinical practice to support anticipatory pain management, external validation in other clinical settings is necessary.

Statistical Forecasts Based on the National Meteorological Center's Numerical Weather Prediction System

The production of interpretive weather element forecasts from dynamical model output variables is now an integral part of the centralized guidance systems of weather services throughout the world. The statistical forecasting system in the United States probably generates the most extensive suite of operational products, although other nations including Australia, Canada, France, Italy, The Netherlands, and the United Kingdom also routinely provide guidance for many weather elements and locations. The United States' statistical guidance system has evolved throughout the past 20 yr. The two principal formulation methods that have been employed are the model output statistics (MOS) and “perfect prog” approaches. These techniques have advantages and disadvantages that influence both aggregate and specific day-to-day performance characteristics of the associated weather element forecasts. Verification results indicate that forecasts from both statistical approaches provide useful guidance for most weather elements and projections for locations throughout the contiguous United States and Alaska. The MOS forecasts have generally been superior to the perfect prog guidance; the drawback to MOS is the necessity to rely on a relatively stable numerical prediction model. As dynamical models change and increase in skill, the perfect prog approach may be preferred for some applications.