Towards cost-sensitive adaptation: when is it worth updating your predictive model?

Our digital universe is rapidly expanding,more and more daily activities are digitally recorded, data arrives in streams, it needs to be analyzed in real time and may evolve over time. In the last decade many adaptive learning algorithms and prediction systems, which can automatically update themselves with the new incoming data, have been developed. The majority of those algorithms focus on improving the predictive performance and assume that model update is always desired as soon as possible and as frequently as possible. In this study we consider potential model update as an investment decision, which, as in the financial markets, should be taken only if a certain return on investment is expected. We introduce and motivate a new research problem for data streams ? cost-sensitive adaptation. We propose a reference framework for analyzing adaptation strategies in terms of costs and benefits. Our framework allows to characterize and decompose the costs of model updates, and to asses and interpret the gains in performance due to model adaptation for a given learning algorithm on a given prediction task. Our proof-of-concept experiment demonstrates how the framework can aid in analyzing and managing adaptation decisions in the chemical industry.

On the predictive content of leading indicators: the case of U.S. real estate markets

This paper employs a probit and a Markov switching model using information from the Conference Board Leading Indicator and other predictor variables to forecast the signs of future rental growth in four key U.S. commercial rent series. We find that both approaches have considerable power to predict changes in the direction of commercial rents up to two years ahead, exhibiting strong improvements over a naïve model, especially for the warehouse and apartment sectors. We find that while the Markov switching model appears to be more successful, it lags behind actual turnarounds in market outcomes whereas the probit is able to detect whether rental growth will be positive or negative several quarters ahead.

Drift from the use of handheld knapsack pesticide sprayers in Boyacá (Colombian Andes)

Offsite pesticide losses in tropical mountainous regions have been little studied. One example is measuring pesticide drift soil deposition, which can support pesticide risk assessment for surface water, soil, bystanders, off target plants and fauna. This is considered a serious gap, given the evidence of pesticide-related poisoning in those regions. Empirical data of drift deposition of a pesticide surrogate, Uranine tracer, within one of the highest potato producing regions in Colombia, characterized by small plots and mountain orography, is presented. High drift values encountered in our study reflect the actual spray conditions using handled knapsack sprayers. Comparison between measured and predicted drift values using three existing empirical equations showed important underestimation. However, after their optimization based on measured drift information, the equations showed a strong predictive power for this study area and the study conditions. The most suitable curve to assess mean relative drift was the IMAG calculator after optimization.

Testing projected wild bee distributions in agricultural habitats: predictive power depends on species traits and habitat type

Species distribution models (SDM) are increasingly used to understand the factors that regulate variation in biodiversity patterns and to help plan conservation strategies. However, these models are rarely validated with independently collected data and it is unclear whether SDM performance is maintained across distinct habitats and for species with different functional traits. Highly mobile species, such as bees, can be particularly challenging to model. Here, we use independent sets of occurrence data collected systematically in several agricultural habitats to test how the predictive performance of SDMs for wild bee species depends on species traits, habitat type, and sampling technique. We used a species distribution modeling approach parametrized for the Netherlands, with presence records from 1990 to 2010 for 193 Dutch wild bees. For each species, we built a Maxent model based on 13 climate and landscape variables. We tested the predictive performance of the SDMs with independent datasets collected from orchards and arable fields across the Netherlands from 2010 to 2013, using transect surveys or pan traps. Model predictive performance depended on species traits and habitat type. Occurrence of bee species specialized in habitat and diet was better predicted than generalist bees. Predictions of habitat suitability were also more precise for habitats that are temporally more stable (orchards) than for habitats that suffer regular alterations (arable), particularly for small, solitary bees. As a conservation tool, SDMs are best suited to modeling rarer, specialist species than more generalist and will work best in long-term stable habitats. The variability of complex, short-term habitats is difficult to capture in such models and historical land use generally has low thematic resolution. To improve SDMs’ usefulness, models require explanatory variables and collection data that include detailed landscape characteristics, for example, variability of crops and flower availability. Additionally, testing SDMs with field surveys should involve multiple collection techniques.

Towards predictive understanding of regional climate change

Regional information on climate change is urgently needed but often deemed unreliable. To achieve credible regional climate projections, it is essential to understand underlying physical processes, reduce model biases and evaluate their impact on projections, and adequately account for internal variability. In the tropics, where atmospheric internal variability is small compared with the forced change, advancing our understanding of the coupling between long-term changes in upper-ocean temperature and the atmospheric circulation will help most to narrow the uncertainty. In the extratropics, relatively large internal variability introduces substantial uncertainty, while exacerbating risks associated with extreme events. Large ensemble simulations are essential to estimate the probabilistic distribution of climate change on regional scales. Regional models inherit atmospheric circulation uncertainty from global models and do not automatically solve the problem of regional climate change. We conclude that the current priority is to understand and reduce uncertainties on scales greater than 100 km to aid assessments at finer scales.

Equations to predict methane emissions from cows fed at maintenance energy level in pasture-based systems

Ruminant production is a vital part of food industry but it raises environmental concerns, partly due to the associated methane outputs. Efficient methane mitigation and estimation of emissions from ruminants requires accurate prediction tools. Equations recommended by international organizations or scientific studies have been developed with animals fed conserved forages and concentrates and may be used with caution for grazing cattle. The aim of the current study was to develop prediction equations with animals fed fresh grass in order to be more suitable to pasture-based systems and for animals at lower feeding levels. A study with 25 nonpregnant nonlactating cows fed solely fresh-cut grass at maintenance energy level was performed over two consecutive grazing seasons. Grass of broad feeding quality, due to contrasting harvest dates, maturity, fertilisation and grass varieties, from eight swards was offered. Cows were offered the experimental diets for at least 2 weeks before housed in calorimetric chambers over 3 consecutive days with feed intake measurements and total urine and faeces collections performed daily. Methane emissions were measured over the last 2 days. Prediction models were developed from 100 3-day averaged records. Internal validation of these equations, and those recommended in literature, was performed. The existing in greenhouse gas inventories models under-estimated methane emissions from animals fed fresh-cut grass at maintenance while the new models, using the same predictors, improved prediction accuracy. Error in methane outputs prediction was decreased when grass nutrient, metabolisable energy and digestible organic matter concentrations were added as predictors to equations already containing dry matter or energy intakes, possibly because they explain feed digestibility and the type of energy-supplying nutrients more efficiently. Predictions based on readily available farm-level data, such as liveweight and grass nutrient concentrations were also generated and performed satisfactorily. New models may be recommended for predictions of methane emissions from grazing cattle at maintenance or low feeding levels.

Menopausal status: A possible predictive factor for recurrence in women with cancer of the uterine cervix without pelvic lymph node metastasis

Objectives: To evaluate risk factors for recurrence of carcinoma of the uterine cervix among women who had undergone radical hysterectomy without pelvic lymph node metastasis, while taking into consideration not only the classical histopathological factors but also sociodemographic, clinical and treatment-related factors. Study design: This was an exploratory analysis on 233 women with carcinoma of the uterine cervix (stages IB and IIA) who were treated by means of radical hysterectomy and pelvic lymphadenectomy, with free surgical margins and without lymph node metastases on conventional histopathological examination. Women with histologically normal lymph nodes but with micrometastases in the immunohistochemical analysis (AE1/AE3) were excluded. Disease-free survival for sociodemographic, clinical and histopathological variables was calculated using the Kaplan-Meier method. The Cox proportional hazards model was used to identify the independent risk factors for recurrence. Results: Twenty-seven recurrences were recorded (11.6%), of which 18 were pelvic, four were distant, four were pelvic + distant and one was of unknown location. The five-year disease-free survival rate among the study population was 88.4%. The independent risk factors for recurrence in the multivariate analysis were: postmenopausal status (HR 14.1; 95% CI: 3.7-53.6; P < 0.001), absence of or slight inflammatory reaction (HR 7.9; 95% CI: 1.7-36.5; P = 0.008) and invasion of the deepest third of the cervix (FIR 6.1; 95% CI: 1.3-29.1; P = 0.021). Postoperative radiotherapy was identified as a protective factor against recurrence (HR 0.02; 95% CI: 0.001-0.25; P = 0.003). Conclusion: Postmenopausal status is a possible independent risk factor for recurrence even when adjusted for classical prognostic factors (such as tumour size, depth of turnout invasion, capillary embolisation) and treatment-related factors (period of treatment and postoperative radiotherapy status). (C) 2009 Elsevier Ireland Ltd. All rights reserved.

Use of a predictive model for food insecurity estimates in Brazil

In 2004 the National Household Survey (Pesquisa Nacional par Amostras de Domicilios - PNAD) estimated the prevalence of food and nutrition insecurity in Brazil. However, PNAD data cannot be disaggregated at the municipal level. The objective of this study was to build a statistical model to predict severe food insecurity for Brazilian municipalities based on the PNAD dataset. Exclusion criteria were: incomplete food security data (19.30%); informants younger than 18 years old (0.07%); collective households (0.05%); households headed by indigenous persons (0.19%). The modeling was carried out in three stages, beginning with the selection of variables related to food insecurity using univariate logistic regression. The variables chosen to construct the municipal estimates were selected from those included in PNAD as well as the 2000 Census. Multivariate logistic regression was then initiated, removing the non-significant variables with odds ratios adjusted by multiple logistic regression. The Wald Test was applied to check the significance of the coefficients in the logistic equation. The final model included the variables: per capita income; years of schooling; race and gender of the household head; urban or rural residence; access to public water supply; presence of children; total number of household inhabitants and state of residence. The adequacy of the model was tested using the Hosmer-Lemeshow test (p=0.561) and ROC curve (area=0.823). Tests indicated that the model has strong predictive power and can be used to determine household food insecurity in Brazilian municipalities, suggesting that similar predictive models may be useful tools in other Latin American countries.

Densities and Viscosities of vegetable oils of nutritional value

Densities and viscosities of five vegetable oils (Babassu oil, Buriti oil, Brazil nut oil, macadamia oil, and grape seed oil) and of three blends of Buriti oil and soybean oil were measured as a function of temperature and correlated by empirical equations. The estimation capability of two types of predictive methodologies was tested using the measured data. The first group of methods was based on the fatty acid composition of the oils, while the other was based on their triacylglycerol composition, as a multicomponent system. In general, the six models tested presented a good representation of the physical properties considered in this work. A simple method of calculation is also proposed to predict the dynamic viscosity of methyl and ethyl ester biodiesels, based on the fatty acid composition of the original oil. Data presented in this work and the developed model can be valuable for designing processes and equipment for the edible oil industry and for biodiesel production.

A Survey on Graphical Methods for Classification Predictive Performance Evaluation

Predictive performance evaluation is a fundamental issue in design, development, and deployment of classification systems. As predictive performance evaluation is a multidimensional problem, single scalar summaries such as error rate, although quite convenient due to its simplicity, can seldom evaluate all the aspects that a complete and reliable evaluation must consider. Due to this, various graphical performance evaluation methods are increasingly drawing the attention of machine learning, data mining, and pattern recognition communities. The main advantage of these types of methods resides in their ability to depict the trade-offs between evaluation aspects in a multidimensional space rather than reducing these aspects to an arbitrarily chosen (and often biased) single scalar measure. Furthermore, to appropriately select a suitable graphical method for a given task, it is crucial to identify its strengths and weaknesses. This paper surveys various graphical methods often used for predictive performance evaluation. By presenting these methods in the same framework, we hope this paper may shed some light on deciding which methods are more suitable to use in different situations.

INDEFINITE QUASILINEAR ELLIPTIC EQUATIONS IN EXTERIOR DOMAINS WITH EXPONENTIAL CRITICAL GROWTH

This paper is concerned with the existence of solutions for the quasilinear problem {-div(vertical bar del u vertical bar(N-2) del u) + vertical bar u vertical bar(N-2) u = a(x)g(u) in Omega u = 0 on partial derivative Omega, where Omega subset of R(N) (N >= 2) is an exterior domain; that is, Omega = R(N)\omega, where omega subset of R(N) is a bounded domain, the nonlinearity g(u) has an exponential critical growth at infinity and a(x) is a continuous function and changes sign in Omega. A variational method is applied to establish the existence of a nontrivial solution for the above problem.