No abstract

NED-2 is a Windows-based system designed to improve project-level planning and decision making by providing useful and scientifically sound information to natural resource managers. Resources currently addressed include visual quality, ecology, forest health, timber, water, and wildlife. NED-2 expands on previous versions of NED applications by integrating treatment prescriptions, growth simulation, and alternative comparisons with evaluations of multiple resources across a management unit. The NED-2 system is adaptable for small private holdings, large public properties, or cooperative managernent across multiple ownerships. NED-2 implements a goal-driven decision process that ensures that all relevant goals are considered; the character and current condition of forestland are known; alternatives to manage the land are designed and tested; the future forest under each alternative is simulated; and the alternative selected achieves the owner's goals. NED-2 is designed to link with ' The computer programs described in this document are available with the understanding that the U.S. Department of Agriculture cannot assure their accuracy, completeness, reliability, or suitability for any purposes other than that reported. The use of trade, firm, or corporation names in this publication is for the infomation and convenience of the reader. Such use does not constitute an official endorsement or approval by the U.

Predicting alfalfa biomass and crop yield for livestock feed is important to the daily lives of virtually everyone, and many features of data from this domain combined with corresponding weather data can be used to train machine learning models for yield prediction. In this work, we used yield data of different alfalfa varieties from multiple years in Kentucky and Georgia, and we compared the impact of different feature selection methods on machine learning (ML) models trained to predict alfalfa yield. Linear regression, regression trees, support vector machines, neural networks, Bayesian regression, and nearest neighbors were all developed with cross validation. The features used included weather data, historical yield data, and the sown date. The feature selection methods that were compared included a correlation-based method, the ReliefF method, and a wrapper method. We found that the best method was the correlation-based method, and the feature set it found consisted of the Julian day of the harvest, the number of days between the sown and harvest dates, cumulative solar radiation since the previous harvest, and cumulative rainfall since the previous harvest. Using these features, the k-nearest neighbor and random forest methods achieved an average R value over 0.95, and average mean absolute error less than 200 lbs./acre. Our top R2 of 0.90 beats a previous work’s best R2 of 0.87. Our primary contribution is the demonstration that ML, with feature selection, shows promise in predicting crop yields even on simple datasets with a handful of features, and that reporting accuracies in R and R2 offers an intuitive way to compare results among various crops.

No abstract

The Web Ontology Language OWL is currently the most prominent formalism for representing ontologies in Semantic Web applications. OWL is based on description logics, and automated reasoners are used to infer knowledge implicitly present in OWL ontologies. However, because typical description logics obey the classical principle of explosion, reasoning over inconsistent ontologies is impossible in OWL. This is so despite the fact that inconsistencies are bound to occur in many realistic cases, e.g., when multiple ontologies are merged or when ontologies are created by machine learning or data mining tools. In this paper, we present four-valued paraconsistent description logics which can reason over inconsistencies. We focus on logics corresponding to OWL DL and its profiles. We present the logic SROIQ4, showing that it is both sound relative to classical SROIQ and that its embedding into SROIQ is consequence preserving. We also examine paraconsistent varieties of EL ++ , DL-Lite, and Horn-DLs. The general framework described here has the distinct advantage of allowing classical reasoners to draw sound but nontrivial conclusions from even inconsistent knowledge bases. Truth-value gaps and gluts can also be selectively eliminated from models (by inserting additional axioms into knowledge bases). If gaps but not gluts are eliminated, additional classical conclusions can be drawn without affecting paraconsistency.