RESUMO
Characterization of smallholder farmers has been conducted in various researches by using machine learning algorithms, participatory and expert-based methods. All approaches used end up with the development of some subgroups known as farm typologies. The main purpose of this paper is to highlight the main approaches used to characterize smallholder farmers, presenting the pros and cons of the approaches. By understanding the nature and key advantages of the reviewed approaches, the paper recommends a hybrid approach towards having predictive farm typologies. Search of relevant research articles published between 2007 and 2018 was done on ScienceDirect and Google Scholar. By using a generated search query, 20 research articles related to characterization of smallholder farmers were retained. Cluster-based algorithms appeared to be the mostly used in characterizing smallholder farmers. However, being highly unpredictable and inconsistent, use of clustering methods calls in for a discussion on how well the developed farm typologies can be used to predict future trends of the farmers. A thorough discussion is presented and recommends use of supervised models to validate unsupervised models. In order to achieve predictive farm typologies, three stages in characterization are recommended as tested in smallholder dairy farmers datasets: (a) develop farm types from a comparative analysis of more than two unsupervised learning algorithms by using training models, (b) assess the training models' robustness in predicting farm types for a testing dataset, and (c) assess the predictive power of the developed farm types from each algorithm by predicting the trend of several response variables.
RESUMO
The heterogeneity of smallholder dairy production systems complicates service provision, information sharing, and dissemination of new technologies, especially those needed to maximize productivity and profitability. In order to obtain homogenous groups within which interventions can be made, it is necessary to define clusters of farmers who undertake similar management activities. This paper explores robustness of production cluster definition using various unsupervised learning algorithms to assess the best approach to define clusters. Data were collected from 8179 smallholder dairy farms in Ethiopia and Tanzania. From a total of 500 variables, selection of the 35 variables used in defining production clusters and household membership to these clusters was determined by Principal Component Analysis and domain expert knowledge. Three clustering algorithms, K-means, fuzzy, and Self-Organizing Maps (SOM), were compared in terms of their grouping consistency and prediction accuracy. The model with the least household reallocation between clusters for training and testing data was deemed the most robust. Prediction accuracy was obtained by fitting a model with fixed effects model including production clusters on milk yield, sales, and choice of breeding method. Results indicated that, for the Ethiopian dataset, clusters derived from the fuzzy algorithm had the highest predictive power (77% for milk yield and 48% for milk sales), while for the Tanzania data, clusters derived from Self-Organizing Maps were the best performing. The average cluster membership reallocation was 15%, 12%, and 34% for K-means, SOM, and fuzzy, respectively, for households in Ethiopia. Based on the divergent performance of the various algorithms evaluated, it is evident that, despite similar information being available for the study populations, the uniqueness of the data from each country provided an over-riding influence on cluster robustness and prediction accuracy. The results obtained in this study demonstrate the difficulty of generalizing model application and use across countries and production systems, despite seemingly similar information being collected.
