Application of fuzzy cognitive maps for cotton yield management in precision farming

  • Authors:

  • Elpiniki I. Papageorgiou;Athanasios Markinos;Theofanis Gemptos

  • Affiliations:

  • Department of Informatics and Computer Technology, Technological Educational Institute (TEI) of Lamia, 3rd Km Old National Road, Lamia-Athens, 35100 LAMIA, Greece;Laboratory of Farm Mechanization, Department of Agriculture, Crop Production and Rural Environment, University of Thessaly, Greece;Laboratory of Farm Mechanization, Department of Agriculture, Crop Production and Rural Environment, University of Thessaly, Greece

  • Venue:
  • Expert Systems with Applications: An International Journal
  • Year:
  • 2009

Quantified Score

Hi-index 12.06

Visualization

Abstract

The management of cotton yield behavior in agricultural areas is a very important task because it influences and specifies the cotton yield production. An efficient knowledge-based approach utilizing the method of fuzzy cognitive maps (FCMs) for characterizing cotton yield behavior is presented in this research work. FCM is a modelling approach based on exploiting knowledge and experience. The novelty of the method is based on the use of the soft computing method of fuzzy cognitive maps to handle experts' knowledge and on the unsupervised learning algorithm for FCMs to assess measurement data and update initial knowledge. The advent of precision farming generates data which, because of their type and complexity, are not efficiently analyzed by traditional methods. The FCM technique has been proved from the literature efficient and flexible to handle experts' knowledge and through the appropriate learning algorithms can update the initial knowledge. The FCM model developed consists of nodes linked by directed edges, where the nodes represent the main factors in cotton crop production such as texture, organic matter, pH, K, P, Mg, N, Ca, Na and cotton yield, and the directed edges show the cause-effect (weighted) relationships between the soil properties and cotton field. The proposed method was evaluated for 360 cases measured for three subsequent years (2001, 2003 and 2006) in a 5ha experimental cotton yield. The proposed FCM model enhanced by the unsupervised nonlinear Hebbian learning algorithm, was achieved a success of 75.55%, 68.86% and 71.32%, respectively for the years referred, in estimating/predicting the yield between two possible categories (''low'' and ''high''). The main advantage of this approach is the sufficient interpretability and transparency of the proposed FCM model, which make it a convenient consulting tool in describing cotton yield behavior.