FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W830223884> ?p ?o ?g. }

• W830223884 abstract "Negative data is defined by observations of unsuccessful events or poor performance. Traditional wisdom dictates that negative data be eliminated from training data sets. This paper presents three step method for incorporating negative data into the rule induction process. The first step is to deploy rule induction using data set containing only positive data. This is traditionally how rule induction techniques such as ID3, C4.5 and CART are used. The second step is to create training data set that contains all of the positive data from Step 1 and also incorporates negative data. The dependent variable from Step 1 becomes dependent variable in the new data set, and new performance-related independent variable is defined. Decision rules are generated using the same rule induction algorithm used in Step 1. The third and final step is to reconcile the two rule sets. A step-wise procedure for creating final, robust rule set is proposed. An example application, related to Just-In-Time manufacturing, is presented in which decision rules are generated using the classification and regression tree (CART) technique. Introduction Inductive reasoning starts with observed data and cases and ultimately generalizes from them to build new rules. These rules are a natural vehicle for what we take to be the most fundamental learning mechanism: prediction-based evaluation of the knowledge store (Holland et al., 1986). One of the critical challenges in learning set of rules is to derive small number of robust rules. While it is possible to derive rules from successes (positive data) and failures (negative data), traditional wisdom dictated that negative information be eliminated from the training data set. However, in many environments it may be desirable to learn from an archived history of data that contains negative information (Triantaphyllou and Soyster, 1996) (Hall, Hansen and Lang, 1997). The purpose of this paper is to present method that supports inductive learning in that (1) can accurately classify and predict successful and unsuccessful performance and (2) can reconcile rules generated from training sets with just positive data and rules generated with both positive and negative data. The development of algorithms for rule induction began with Hunt’s Concept Learning System (Hunt et al., 1966) and was followed by Quinlan’s ID3 algorithm (Quinlan, 1979). In 1984 Breiman, Friedman, Olshen and Stone (1984) developed nonparametric statistical procedure, classification and regression trees (CART), to analyze categorical and continuous data using exhaustive searches and computer intensive testing to select an optimal decision tree. Crawford (1989) states that in cases where data is noisy, CART is a remarkably sophisticated tool for concept induction. Inductive learning techniques typically generate decision rules by training on data sets that contain only positive data. However, there are wide variety of reasons for wanting to learn from data representing less than optimal conditions. First, it is important to learn from mistakes. The precise conditions that caused poor performance can be identified and steps can be taken to rectify the situation in the future. Secondly, by incorporating negative data with the positive data it increases the number of observations in the training set. As result, more robust classifiers can be constructed. Finally, by analyzing both good and poor performance it is possible for the analyst to uncover the predictive structure of the problem. This means that the relationship between the variables that cause negative (poor) performance can be discovered and measures to assure positive (good) performance can then be taken. An Application of the Three Step Method to JIT Manufacturing In Just-In-Time (JIT) manufacturing, the kanban is visual cue that is used to signal the replenishment of goods at each stage in the production process. The number of circulating kanbans is important to the effective operation of the JIT production system. Too many kanban cards produce excess work-in-progress inventory, while too few lead to production-floor disturbances. Moreover, the number of kanbans can significantly influence the load balance between processes, and the amount of orders needed to obtain supplies from subcontractors. Quite often JIT with kanbans is used in environments not meeting the conditions for optimal performance. These conditions may be unstable product demand, highly variable processing times, or highly variable vendor supply times. When these conditions exist buffer of inventory is necessary to smooth production flow in the shop. The result is factory “bloated” with work-in-progress inventory often characterized by large number of kanbans at each" @default.
• W830223884 created "2016-06-24" @default.
• W830223884 creator A5003627417 @default.
• W830223884 creator A5012377065 @default.
• W830223884 creator A5022259270 @default.
• W830223884 date "1998-01-01" @default.
• W830223884 modified "2023-09-25" @default.
• W830223884 title "A Method for Incorporating Negative Data into Rule Induction" @default.
• W830223884 cites W114712355 @default.
• W830223884 cites W1594031697 @default.
• W830223884 cites W2056218787 @default.
• W830223884 cites W2075857542 @default.
• W830223884 cites W2293077137 @default.
• W830223884 hasPublicationYear "1998" @default.
• W830223884 type Work @default.
• W830223884 sameAs 830223884 @default.
• W830223884 citedByCount "1" @default.
• W830223884 crossrefType "journal-article" @default.
• W830223884 hasAuthorship W830223884A5003627417 @default.
• W830223884 hasAuthorship W830223884A5012377065 @default.
• W830223884 hasAuthorship W830223884A5022259270 @default.
• W830223884 hasConcept C111919701 @default.
• W830223884 hasConcept C119857082 @default.
• W830223884 hasConcept C124101348 @default.
• W830223884 hasConcept C134306372 @default.
• W830223884 hasConcept C149271511 @default.
• W830223884 hasConcept C154945302 @default.
• W830223884 hasConcept C177264268 @default.
• W830223884 hasConcept C182365436 @default.
• W830223884 hasConcept C199360897 @default.
• W830223884 hasConcept C2776780472 @default.
• W830223884 hasConcept C33923547 @default.
• W830223884 hasConcept C41008148 @default.
• W830223884 hasConcept C84525736 @default.
• W830223884 hasConcept C84839998 @default.
• W830223884 hasConcept C98045186 @default.
• W830223884 hasConceptScore W830223884C111919701 @default.
• W830223884 hasConceptScore W830223884C119857082 @default.
• W830223884 hasConceptScore W830223884C124101348 @default.
• W830223884 hasConceptScore W830223884C134306372 @default.
• W830223884 hasConceptScore W830223884C149271511 @default.
• W830223884 hasConceptScore W830223884C154945302 @default.
• W830223884 hasConceptScore W830223884C177264268 @default.
• W830223884 hasConceptScore W830223884C182365436 @default.
• W830223884 hasConceptScore W830223884C199360897 @default.
• W830223884 hasConceptScore W830223884C2776780472 @default.
• W830223884 hasConceptScore W830223884C33923547 @default.
• W830223884 hasConceptScore W830223884C41008148 @default.
• W830223884 hasConceptScore W830223884C84525736 @default.
• W830223884 hasConceptScore W830223884C84839998 @default.
• W830223884 hasConceptScore W830223884C98045186 @default.
• W830223884 hasLocation W8302238841 @default.
• W830223884 hasOpenAccess W830223884 @default.
• W830223884 hasPrimaryLocation W8302238841 @default.
• W830223884 hasRelatedWork W104698897 @default.
• W830223884 hasRelatedWork W110229806 @default.
• W830223884 hasRelatedWork W1429952601 @default.
• W830223884 hasRelatedWork W1530582268 @default.
• W830223884 hasRelatedWork W1557512086 @default.
• W830223884 hasRelatedWork W1599651760 @default.
• W830223884 hasRelatedWork W2016071450 @default.
• W830223884 hasRelatedWork W2123194680 @default.
• W830223884 hasRelatedWork W2167365327 @default.
• W830223884 hasRelatedWork W2170804782 @default.
• W830223884 hasRelatedWork W2253738698 @default.
• W830223884 hasRelatedWork W2355855810 @default.
• W830223884 hasRelatedWork W2487602329 @default.
• W830223884 hasRelatedWork W2564526664 @default.
• W830223884 hasRelatedWork W2766440341 @default.
• W830223884 hasRelatedWork W2777570032 @default.
• W830223884 hasRelatedWork W2973288386 @default.
• W830223884 hasRelatedWork W3138557076 @default.
• W830223884 hasRelatedWork W3140758913 @default.
• W830223884 hasRelatedWork W2557556902 @default.
• W830223884 isParatext "false" @default.
• W830223884 isRetracted "false" @default.
• W830223884 magId "830223884" @default.
• W830223884 workType "article" @default.