Using Fuzzy Logic or Probability Approach in Revising Unknown, Invalid, or Missing Data Points: Application to Shrimp Data Files in the Gulf of Mexico, Years 2005 and 2006

摘要

Probability and fuzzy logic have played important roles in data analysis, prediction, and estimation. Fuzzy logic can help to solve complex problems using approximations and allows users to analyze incomplete and imprecise data sets. Fishery data in general contain missing data points (see for example, [1, 2] or shrimp data files, Gulf of Mexico, 1984 to present). It is of great importance to estimate such points accurately using reliable scientific oriented techniques. The purpose of this study is to introduce the theories of probability and fuzzy logic especially the latter to the shrimp data files for estimating missing/invalid/unknown data points. In the article, these theories along with the statistical mode and multiple imputation were deployed to revise such data points. As an application, shrimp data in the Gulf of Mexico collected by the National Marine Fisheries Service for the years 2005 and 2006 were selected due to the existence of unknown, invalid, or missing values in the species , fathomzone , and subarea fields and the similarity in missing patterns. The methods mentioned above were deployed to revise these fields. The probability approach deployed a discrete multivariate probability distribution developed based on the shrimp data files 2000-2001, statistical mode, and imputation. The fuzzy logic approach also deployed a special form of a Gaussian membership function based on 2000-2001 data files, statistical mode, and imputation. In general, analyses showed that both theories estimated the species , fathomzone , and subarea in a satisfactory manner. However, it was concluded that the fuzzy logic showed more robustness when a large number of data points were to be estimated.