Browsing by Author "Noncheva, Veska"
Now showing 1 - 9 of 9
Results Per Page
Sort Options
- An Approach to Variable Aggregation in Efficiency AnalysisPublication . Noncheva, Veska; Mendes, Armando B.; Silva, EmilianaIn the nonparametric framework of Data Envelopment Analysis the statistical properties of its estimators have been investigated and only asymptotic results are available. For DEA estimators results of practical use have been proved only for the case of one input and one output. However, in the real world problems the production process is usually well described by many variables. In this paper a machine learning approach to variable aggregation based on Canonical Correlation Analysis is presented. This approach is applied for efficiency estimation of all the farms in Terceira Island of the Azorean archipelago.
- Azorean agriculture efficiency by PARPublication . Noncheva, Veska; Mendes, Armando B.; Silva, EmilianaThe producers always aspire at increasing the efficiency of their production process. However, they do not always succeed in optimizing their production. In the last years, the interest on Data Envelopment Analysis (DEA) as a powerful tool for measuring efficiency has increased. This is due to the large amount of data sets collected to better understand the phenomena under study, and, at the same time, to the need of timely and inexpensive information. The “Productivity Analysis with R” (PAR) framework establishes a user-friendly data envelopment analysis environment with special emphasis on variable selection and aggregation, and summarization and interpretation of the results. The starting point is the following R packages: DEA (Diaz-Martinez and Fernandez-Menendez, 2008) and FEAR (Wilson, 2007). The DEA package performs some models of Data Envelopment Analysis presented in (Cooper et al., 2007). FEAR is a software package for computing nonparametric efficiency estimates and testing hypotheses in frontier models. FEAR implements the bootstrap methods described in (Simar and Wilson, 2000). PAR is a software framework using a portfolio of models for efficiency estimation and providing also results explanation functionality. PAR framework has been developed to distinguish between efficient and inefficient observations and to explicitly advise the producers about possibilities for production optimization. PER framework offers several R functions for a reasonable interpretation of the data analysis results and text presentation of the obtained information. The output of an efficiency study with PAR software is self- explanatory. We are applying PAR framework to estimate the efficiency of the agricultural system in Azores (Mendes et al., 2009). All Azorean farms will be clustered into homogeneous groups according to their efficiency measurements to define clusters of “good” practices and cluster of “less good” practices. This makes PAR appropriate to support public policies in agriculture sector in Azores.
- Azorean Agriculture Efficiency by PARPublication . Mendes, Armando B.; Noncheva, Veska; Silva, EmilianaThe producers always aspire at increasing the efficiency of their production process. However, they do not always succeed in optimising their production. In the last years, the interest on Data Envelopment Analysis (DEA) as a powerful tool for measuring efficiency has increased. This is due to the large amount of data sets collected to better understand the phenomena under study and, at the same time, to the need of timely and inexpensive information. The "Productivity Analysis with R" (PAR) framework establishes a user-friendly data envelopment analysis environment with special emphasis on variable selection, aggregation, summarisation and interpretation of the results. The starting point is the following R packages: DEA (Diaz-Martinez and Fernandez-Menendez 2008) and FEAR (Wilson 2008). The DEA package performs some models of data envelopment analysis presented in Cooper et al. (2007). FEAR is a software package for computing nonparametric efficiency estimates and testing hypotheses in frontier models. FEAR implements the bootstrap methods described in Simar and Wilson (2000). PAR is a software framework using a portfolio of models for efficiency estimation and also providing results explanation functionality. PAR framework has been developed to distinguish between efficient and inefficient observations and to explicitly advise the producers about possibilities for production optimisation. PAR framework offers several R functions for a reasonable interpretation of the data analysis results and text presentation of the obtained information. The output of an efficiency study with PAR software is self-explanatory. We are applying PAR framework to estimate the efficiency of the agricultural system in Azores (Mendes et al. 2009). All Azorean farms will be clustered into homogeneous groups according to their efficiency measurements to define clusters of "good" practices and cluster of "less good" practices. This makes PAR appropriate to support public policies in agriculture sector in Azores.
- Canonical correlation analysis and DEA for azorean agriculture efficiencyPublication . Mendes, Armando B.; Noncheva, Veska; Silva, EmilianaIn this paper we will document the application of canonical correlation analysis to variable aggregation using the correlations of the original variables with the canonical variates. A case study, about farms in Terceira Island, with a small data set is presented. In this data set of 30 farms we intend to use 17 input variables and 2 output variables to measure DEA efficiency. Without any data reduction procedure several problems known as “curse of dimensionality” are expected. With the data reduction procedures suggested it was possible to conclude quite acceptable and domain consistent conclusions.
- Canonical Correlation Analysis in Variable Aggregation in DEA.Publication . Mendes, Armando B.; Noncheva, Veska; Silva, EmilianaNeste trabalho documenta-se a aplicação de análise de correlações canónicas à agregação de variáveis em DEA, usando as correlações entre as variáveis originais e os componentes canónicos extraídos. É apresentado um caso de estudo que utiliza um pequeno conjunto de dados sobre explorações agrícolas na ilha terceira. Neste conjunto de 30 explorações agrícolas pretende-se usar 17 variáveis de input e 2 de output para avaliar a eficiência usando DEA. Sem qualquer redução de dados, vários problemas conhecidos como "praga da dimensionalidade" seriam esperados. Com os procedimentos sugeridos foi possível obter resultados razoáveis e de acordo com o conhecimento de domínio actual.
- Decision support for enhanced productivity with R software: an Azorean farms case study.Publication . Mendes, Armando B.; Noncheva, Veska; Silva, EmilianaAzores is a Portuguese insular territory where the main economic activity is dairy and meat farming. Dairy policy depends on Common Agricultural Policy of the European Union and is limited by quotas. On top of that the transformation sector had implemented a program for penalising the worst quality agricultural raw materials. The current historical context is particularly complex as some major changes are likely to occur. This is the case for the increase prices of some food products in international markets and, locally, the end of milk quota system. The multiplying effect of agriculture in both a small economy and the Azorean society, makes of major interest this kind of work not only to protect the income of farmers, but also to keep the society in equilibrium on employment matters and reduce immigration cycles. In this context, decision makers need information and knowledge for deciding the best policies in promoting quality and best practices. So, in this project we apply benchmarketing methodologies to estimate the efficiency of the agricultural system in Azores. We also propose to identify the inefficiency units and delineate action plans for correcting production or organizational identified problems. The data analysis will be possible using non parametric methods like data envelopment analysis – DEA. We develop a new data-driven methodology, called PAR (Productivity Analysis with R), which combines DEA with a statistical technique need for analysing a reduced number of farms. All Terceira (the second biggest island) farms are analyze according to their efficiency measurements to define groups of “good” practices and groups of “less good” practices. This makes the system appropriate to support public policies in agriculture sector in Azores. The decision makers we intend to support are of two different levels: farmers or services responsible for agriculture improvement and political decision makers. These two types of decision makers need information that is very specific and concrete in the first case and much more aggregated and general in the second case. The data analysis methods we are using can support the needs of both decision makers’ types, but the software interface must be specific designed. PAR project is designed to provide a bridge from mathematical models to productivity study using R statistical software. Several DEA models are described in literature. Some of them are implemented as functions in statistical software R which are being used for PAR system. Some works in restricted data sets were already done for the dairy sector in Azores using different approaches, by the authors. We use this data and results to validate and correct the software system we are developing. R statistical software is not very user friendly. Much programing is needed to make the output of the PAR computer program self explanatory and easily understandable.
- Efficient decisions using DEAPublication . Mendes, Armando B.; Silva, Emiliana; Noncheva, VeskaData Envelopment Analysis (DEA) is becoming an increasingly popular management tool for decision support related to efficiency comparisons. The task of the DEA is to evaluate the relative performance of units of a system. Is not a problem solution technique but an important problem analysis method based in mathematical programming with some similarities with Multiple Criteria Decision Analysis (MCDA). DEA makes it possible to identify efficient and inefficient units in a framework where results are considered in their particular context. The units to be assessed should be relatively homogeneous and were originally called Decision Making Units (DMUs). It is an extreme point method that compares each DMU with the "best" DMUs. The “Productivity Analysis with R” (PAR) framework establishes a user - friendly data envelopment analysis environment with special emphasis on variable selection and aggregation, and summarization and interpretation of the results. PAR framework has been developed to distinguish between efficient and inefficient observations of performances and to advise explicitly for producers’ possibilities to optimize their production. In this work we will apply PAR to farms in Terceira Island, with a small data set of 30 farms. This data set includes 14 input variables and 4 output variables. With PAR was possible to conclude that 4 farms are scale – efficient and the others are scale inefficiency due to decreasing returns to scale or increasing returns to scale. This implies that either the dairy farm is too big (the number of cows is too large) or to small and that the farmer can improve the productivity of inputs and hence reduce unit costs by reducing or increasing the dimension of the farm.
- A software framework for measuring efficiency.Publication . Noncheva, Veska; Mendes, Armando B.; Silva, Emiliana"[...]. PAR is a software framework using a variety of models estimating efficiency and providing results explanation functionality. [...]. We are applying PAR framework to estimate the efficiency of the agricultural system in Azores [Mendes et al., 2009]. All Azorean farms will be clustered into homogeneous groups according to their efficiency measurements to define clusters of "good" practices and cluster of "less good" practices. This makes PAR appropriate to support public policies in agriculture sector in Azores. [...]"
- Sustainable Tourism and Agriculture Multifunctionality by PAR: A Variable Selection ApproachPublication . Mendes, Armando B.; Noncheva, Veska; Silva, EmilianaData Envelopment Analysis (DEA) is a popular non-parametric method used to measure efficiency. It uses linear programming to identify points on a convex hull defined by the inputs and outputs of the most efficient Decision Making Units (DMUs). Two critical elements account for the strength of the DEA approach: (1) no a priori structure is placed on the production process of the firm, and (2) the models can yield a measure of efficiency even with a very small number of data points. The first point is particularly important because the measure of efficiency is based upon the best practice of the DMUs at any of the levels of output observed. Data envelopment analysis measures efficiency and is very sensitive to the choice of variables for two reasons: the number of efficient DMUs is directly related to the number of variables, and the selection of the variables greatly affects the measure of efficiency when the number of DMUs is few and/or when the number of explanatory variables needed to compute the measure of efficiency is too large. Our approach advises which variables should be included in a DEA model. Hence, a variable selection method is presented for the deterministic DEA approach. First, a definition of different measures of efficiency and the various DEA models used to measure efficiency is provided, and then a variable selection method is proposed. The Azorean agricultural system is used as an example to illustrate the method.