Heungsun Park

Relative-error prediction

We derive the form of the best mean squared relative error predictor of Y given X. Some methods of estimating predictors with good relative error properties are proposed and studied via simulation. The methods are illustrated with an example in which county-level gasoline sales are predicted from county-level population.

Unbiased estimation of greenhouse whitefly, Trialeurodes vaporariorum, mean density using yellow sticky trap in cherry tomato greenhouses

The spatial distribution of the count of adult greenhouse whiteflies, Trialeurodes vaporariorum (Westwood), on yellow sticky traps was analyzed using Taylors power law and spatial autocorrelation statistics in the cherry tomato greenhouses from 1998–1999. Samples were collected weekly using a cylindrically shaped yellow sticky trap placed in a 5 by 8 grid covering 0.10–0.15 ha in each of five cherry tomato greenhouses. Taylors (1961) power law indicated that counts of T. vaporariorum on traps were aggregated within greenhouses. Spatial autocorrelation analysis showed that trap catches were similar (positively autocorrelated) to a distance of 12.5 m, and then dissimilar (negatively autocorrelated) at >12.5 m. Autocorrelation‐lag plots showed a globally significant spatial relation in 34 of 57 sample‐weeks according to Bonferronis approximation. The presence of this spatial relation was not related to the changes of mean density. Trap counts at the second lag distance (12.5–25 m) showed little spatial autocorrelation and tended to be the most spatially independent. A fixed‐precision‐level sequential sampling plan was developed using the parameters from Taylors power law. The presence of spatial dependency in data sets degraded the sampling plans precision relative to performance in data sets lacking significant spatial autocorrelation. Therefore, to obtain an unbiased mean density of T. vaporariorum per greenhouse, sticky traps should be placed at least >12.5 m apart to ensure that they are spatially independent.

Use of covariates in Taylor’s power law for sequential sampling in pest management

In pest management, the pest density within a group of plants needs to be estimated for taking timely actions, such as spraying pesticides or releasing natural enemies. Taylor’s power law is widely used for identifying the aggregation patterns of the pests and designing a sequential sampling plan to estimate the pest mean density. The conventional estimates given by Taylor’s power law do not consider potential density differences due to various covariates, but focus only on the relationship between the sample means and the variances.In this article, we develop a new sequential sampling stop line based on Taylor’s power law by using quasi-likelihood with covariate effects. The simulation results show that the proposed estimators are better than the conventional estimators in terms of mean squared error. For validation and evaluation of the sampling stop line given by the proposed estimator, we use RVSP software in which actual observations are randomly and iteratively sampled until the total number of a pest reaches the stop line. We demonstrate both types of sequential sampling stop lines, those based on the conventional method and those based on the new method, for the population density of spider mites on glasshouse roses.

A Shrinked Forecast in Stationary Processes Favouring Percentage Error

In stationary time-series forecasting, the commonly used criterion for selecting a proper forecast is the mean square error (MSE), which is minimized by the conditional expectation of future observation given the entire past known as a minimum MSE forecast. In this paper, mean square percentage error (MSPE) instead of is used to forecast autoregressive moving average (ARMA)(p,q) series. The suggested forecast takes the form of or (CV_t+1 is the coefficient of variation for one step ahead) times the minimum MSE forecast, which performs better not only in MSPE, but also in mean absolute percentage error (MAPE) than the ordinary MSE forecast in simulation studies. A real data example also supported this result. We conclude that, if percentage error is a prime concern, this shrinked version of MSE forecast performs better than the ordinary forecast in the stationary ARMA(p,q) model. Copyright 2005 Blackwell Publishing Ltd.

Practical ontology systems for enterprise application

One of the main challenges in building enterprise applications has been to balance between built-in functionality and domain/scenario-specific customization. The lack of formal ways to extract, distill and standardize the embedded domain knowledge has been a barrier to effective and efficient customization. Ontology may provide, as many would hope, the much needed methodology and standard to achieve the objective of building flexible enterprise solutions. This article examines the uses, issues and challenges of using ontology in enterprise applications. We believe that we are seriously lacking in modeling methodology, domain user tools, and lifecycle management methodology for the creation and maintenance of ontology on a large deployable scale. We present the issues based on an ongoing project to build a product ontology for a public procurement system. Through real life scenarios, we are hoping to convey important research directions to better enable ontology.

Estimation of Leafmine Density of Liriomyza trifolii (Diptera: Agromyzidae) in Cherry Tomato Greenhouses using Fixed Precision Sequential Sampling Plans

This study was conducted to develop sequential sampling plans to estimate leafmine density by Liriomyza trifolii (Burgess) at three fixed-precision levels in commercial tomato greenhouses. The within-greenhouse spatial patterns of leafmines were aggregated. The slopes and intercepts of Taylors power law did not differ between greenhouses and years. A fixed-precision level sampling plan was developed using the parameters of Taylors power law generated from total number of leafmines in a tomato leaf (consisted of 7 leaflets) at three precision levels (D) of 0.20, 0.25, and 0.30. The resulting sampling plans were tested with sequential bootstrap simulations (n = 500) using 9 independent data sets for validation. Bootstrap simulation within a wide range of densities demonstrated that actual D′ values at desired D = 0.30 averaged < 0.29 in all cases. even at the lowest density of leafmine (0.16 mines per leaf), the actual mean D′ value was 0.24 at D = 0.30. This result shows that the sampling plan developed in this study is effective and reliable for estimating the mine densities in tomato greenhouses.

A study on RFID/USN based e-pedigree system for cold chain management

Purchase and consumption patterns in the food and pharmaceutical industries have changed because of the increase in fresh, convenient, high-quality foods that are available due to improvements in economic level and changes in lifestyle that are part of modern society. Efficient cold chain management has been receiving increased attention in the logistics industry due to the rapid increase in customer demand for refrigerated and frozen food, and pharmaceutical products. Electronic pedigree (e-Pedigree) has been suggested as a mean of providing transparent and secure product trading histories of pharmaceutical products across the entire US supply chain. It can help preserve peoples health and protect the pharmaceutical industries by preventing distribution of contaminated or counterfeit products. The proposed RFID/USN (radio frequency identification/ubiquitous sensor network) based e-pedigree system is based on the EPCglobal Pedigree Ratified Standard version 1.0 (ePedigree). Unfortunately, this system does not include environmental status information, which is necessary for ensuring the proper conditions during storage, transportation, and distribution of food and pharmaceutical products. The main purpose of this study is to describe an RFID/USN based e-pedigree system for temperature-sensitive foods and pharmaceutical products to guarantee the quality of these products and ensure they do not become contaminated anywhere throughout the supply chain.

The effect of psychological ownership on the possession attachment and willingness to share the Internet content

The development of the Internet has accelerated the speed of information distribution, and lead to the emergence of content sharing among users. In this study, we try to understand content sharing over the public Web using the concept of psychological ownership. The study consists of two different experiments. In the first study, we investigated how the right to control of their digital contents and time/ effort cost are related to willingness to share. In the second experiment, we add the concept of the psychological distance. Our study contributes to the understanding of peoplepsilas information sharing behavior by focusing on psychological ownership and psychological in the public Internet domain.

Application of Sequential Classification of Prey/Predator Ratio Test to Tetranychus urticae and Phytoseiulus persimilis System in Greenhouse Roses

Abstract A sequential sampling procedure for classifying the ratio of prey/predators developed by Nyrop (1988) was examined to implement for a biological control in the greenhouse roses. The procedure was combined with a sequential density classification procedure for use in monitoring a phytophagous mite, Tetranychus urticae Koch, and a phytoseiid predator, Phytoseiulus persimilis Athias-Henriot. This procedure was required four parameters: Means and variances for T. urticae and P. persimilis , correlation coefficient between densities of prey and predator and critical ratio of prey and predator. The parameter values used in this study were 0.725 for the correlation coefficient and 10 for the critical ratio. The variances for each species were estimated using a Taylors power law model. The procedure is proven to be successfully applicable to T. urticae and P. persimilis system in greenhouse roses at two action threshold levels of 5 and 10 T. urticae per three-leaflet leave. The limitations and implementation of this procedure is also discussed.

A weighted sifting method to improve the effectiveness of collaborative filtering

In this paper, we improve the accuracy of the conventional collaborative filtering algorithm by proposing a weighted sifting method. The weighted sifting method preprocesses the given customer data to generate an adjusted customer data which we believe contains less noise than the original one, and thus effectively discriminates the preference weights of items for each customer. We present two alternative calculation methods for weight adjustment, and our experimental evaluation shows that both calculation methods result in better accuracy than traditional collaborative filtering.