Optimum Growth Ensemble in Agroforestry (OGEA)

Abstract

Agroforestry describes the land use management system in which trees or shrubs are grown around or among crops or pastureland.It combines agricultural and forestry technologies to create more diverse, productive, profitable, healthy, and sustainable land-use systems[1]. The treatment combinations of doses, fertilizers, variety of crops and their spacing i. e. geometrical arrangements, canopy manipulations, crop harvest intervals, irrigation schedules etc. are standardized and judged specifically to develop different Agriculture and Forestry Models. Cluster ensemble technique has been proved to be better than any of the traditional clustering algorithms fordiscovering complicated structures in data. Cluster ensembles can provide robust and stable solutions by leveraging the consensusacross multiple clustering results, while averaging out emergent spurious structures that arisedue to the various biases to which each participating algorithm is tuned. In this paper, a cluster ensemble technique for Optimum Growth Ensemble in Agroforestry (OGEA) has been proposed. OGEAaims at improving robustnessand quality of clustering scheme, particularly in Agroforestry sector which in turn enhance the production and productivity of any crop. OGEA consists of four phases. First phase generates the various clustering schemes. This phase does the relabeling to avoid the label correspondence problem. The second phase predicts the tuples by using the three different techniques of prediction viz., Discriminant Analysis, Multilayer perceptron and Logistic regression. In the phase III, depending upon the results of the best technique and threshold of the consensus function obtained by various clustering schemes, consensus partition is generated. In the phase IV, Performance Groups are determined in descending order of optimum resultsi. e. Performance Group 1 gives the maximum yield or survival percentage followed by other Performance Groups respectively. Extensive experimentation has been done on the data setby varying the number of partitions and clusters in cluster ensemble. Different Performance Groups are achieved by using this technique that segregates the various treatment combinations in order to achieve the optimum production. Furthermore, we investigate in depth the about the quality, accuracy and stability of results by using different Performance Groups by utilizing the various quality measures viz., Purity, Normalized Mutual Information (NMI) andAdjusted Rand Index (ARI). Further, the result is statistically tested by determining the comparison of each Performance Group with Control by using the various statistical measures such as Mean, Standard Deviation and Coefficient of Variation.