Mapping Of Provincial Food Security In Indonesia Using Based Clustering Model
Abstract
Indonesia was known as an agrarian and maritime country, should not experience difficulties in fulfill food needs or having high food security. However, it is a formidable challenge for the Indonesiato meeting food needs. The low level of food security was caused more by Indonesia's geographical conditions in the form of islands that cause inequality of food production, distribution and absorption among provinces in Indonesia. To reduce the occurrence of food security inequality between provinces in Indonesia, clusters was formed based on food security indicators. Based clustering technique is chosen because it has feature that can overcome the problem of overlapping. The results of research produce 3 clusters based on the classification of food security levels. The first cluster consisted of 19 provinces with a classification of middle food security levels, the second cluster consisted of 10 provinces with the classification of the level of high food security, and the third cluster consisted of 5 provinces with a classification of low food security levels. It is expected that the results of this clustering can provide input to the Indonesian government to focus more on 5 Provinces with low food security classification, which focuses on access to food.
Introduction
Food is the most basic human needs, so its availability for community must be guaranteed and it is an obligation of a country, including Indonesia. As mandated in the 1945 Constitution of Republic of Indonesia, the State is obliged to carry out food sovereignty (food right) and fulfill population food need. Indonesia is well known as an agrarian and maritime country, should not experiance difficulties in fulfill food need for its population or should have high food security. However, fulfillness of food needs for Indonesia which consists of 34 Provinces is a big challenge. It’s shown from data of Global Hunger Index that released by IFPRI (International Food Policy Research Institute) in 2017. Based on IFPRI data, Indonesia was catagorized country that experiance serious hunger with value of global hunger index of 22. This condition shows that Indonesia’s food security is very low. The serious level of hunger and low food security in Indonesia caused by several factors, one of them is Indonesia’s geographical condition which in the form of islands. This condition causes inequality of production, distribution, and absorption of food between provinces in Indonesia due to the differences of potential natural resource, human resource and level of development between provinces in Indonesia.
The first step to reduce the occurrence of food security inequality between provinces in Indonesia is mapping the state of food security each province in Indonesia. By knowing the state of food security each province in Indonesia, it will be known which provinces are categorized priority provinces to have immediate policy to increase the level of food security implemented. One of the way that can be done to mapping the state of food security by determine tha state of food security of each provinces in Indonesia is forming cluster to classify the province in Indonesia based on the food security indicators. One of method to forming cluser of food security by provinces in Indonesia is based clustering model. It is chosen because its has feature that can overcome overlapping problems that often accur in other clustering methods.
This study aims to mapping the condition of food security in each province in Indonesia. While the benefits of this research are giving the information for the steakholder that can be used as basic planning of policy in effort to increase the food security or reduce the risk of food insecurity in Indoneisa.
Research Method
This study uses secondary data with reference to data year 2017 from literature books, economic and business journals, and publications from Statistics Indonesia. While the analysis method used is clustering method to classify 34 Provinces in Indonesia according to their food security state. This clustring is based on the dimensions of food security variabels from simplification definition of food security proposed by USAID (United States Agency for International Development).
USAID defines food security as a condition when all people at all times have physical and economic access to obtain their consumption needs for a healthy and productive life. Based on this definition, food security has a broad and complex concept influenced by interactions of various agro-physical, socio-economic and biological factors. Like the concept of health or social welfare, there is no definite measure in measuring food security. However, the complexity of measuring food security can be simplified by focusing calculations in three dimensions, namely:
- Food Availability is reached when the amount of food is sufficient and is available consistently for all individuals in a country. This food can be fulfilled through domestic production, imports and food aid. In this study, ratio of food crops availability to normative consumption per capita is used to represent the dimensions of food availability.
- Food Access, it is guaranteed when the household and all individuals in it have adequate resources to obtain nutritious food. This access depends on household income, household income distribution and food prices.
- Food Utilization, it is the use of biologically appropriate food, which makes a food meet important energy and nutrition needs, drinking water, and adequate sanitation. Effective use of food depends on how broad is household knowledge in storage and good food processing techniques, basic principles of proper nutrition and child care, and disease management. Prior to grouping provinces based on their food security status the first thing to do is to determine indicators in the formation of provincial clusters in Indonesia according to their food security state. This needs to be done because not all indicators can describe the condition of food security in a region with good and have complete availability of data. Based on the differences in the availability and the completeness of the region's data, this study use six indicators that have been identified and compiled to determine the status of provincial food security in Indonesia. All of these indicators are considered to represent the dimensions of food security according to USAID and FAO (Food and Agriculture Organization).
Based clustering model was introduced by Banfield and Raftery (1993) for grouping objects in a population that describes a clustering approach, where a group of populations is identified based on probability distributions and the entire population is modeled as a distribution mixture. Based clustering model uses the assumption that in a population sub-populations can be taken which have a certain probability distribution and each sub-population has different parameters. In the finite mixture model there are at least two main processes that need to be done, namely parameter estimation and selection of the best model that describes the data structure. The choice of the best model can be done using the likelihood approach and the bayesian approach in the finite mixture model for clustering, the best model that describes the data structure will be obtained. Bayesian Information Criterion (BIC) is a measure that can be used in the selection of models in this study. The use of Bayesian Information Criterion (BIC) in this study is because BIC has a more robust nature. The best model is the model that produces the largest BIC value. After obtaining the best model, parameter estimation and clustering of research objects were carried out, research objects in this study were 34 provinces in Indonesia.
The measurement of food security classification is done by comparing the cluster mean with confidence interval (95%) of the variables used in the study. If the mean cluster is below the lower limit of the confidence interval then the classification of food security achievement is low, if the mean cluster between the confidence intervals then the classification of food security achievement is intermediate, and if the mean cluster is upper limited to the confidence interval then the classification of the class is high food security achievements.
Discussion
The first step that needs to be done to form a cluster with based clustering model is choosing the best model. The selection of the best model in this study is done by Bayesian Information Criterion. Using Mclust 5. 4. 1 software, this calculation produced 9 models, which each model was tried using the number of clusters ranging from 1-9 clusters. The largest BIC value indicates that the model used with a certain number of clusters is the most feasible model or the best model. The processing results recommend three best models with optimal number of clusters with each BIC value.
The best model according to BIC criteria is the EVI model with the number of clusters as much as 3. the EVI model is a cluster model with diagonal distribution, the volume is Equal (same), the shape is variable (different) and its orientation is coordinate axes. BIC criteria also becomes basis for obtaining the number of optimal culsters and members of each clusters.
After getting the best model, the next step to do is estimating the parameter. Parameter estimation using EM algorithm. Model based clustering produces several parameters including mixing probabilities (π), mean vectors (µ) and matrix variance covariance (Σ).
Mixing probabilities show cluster membership degree where cluster 1 is 56% or 19 provinces, cluster 2 is 29% or as many as 10 provinces and cluster 3 is 15% or as many as 5 provinces. To measure the classification of provincial food security status in Indonesia in each cluster, it is done by comparing the group mean formed with a 95% confidence interval calculated from the mean and variable standard deviations. There are two types of variables used in this study, namely positive and negative variables. A positive variable is a variable that, if the value is greater, indicates a good condition, while a negative variable is a variable which, if the value is greater, indicates an unfavorable state. For positive variables, if the mean cluster is below the confidence interval, the classification of food security achievement is low, the mean cluster between the confidence interval of medium food security achievement and if the mean cluster is above the confidence interval, the food security achievement is high. The opposite applies to negative variables.
Positive variables used in this study were ratio of food crops availability to normative consumption per capita (X1), GRDP (Gross Regional Domestic Product) per capita (X3), life expectancy (X4), average consumption of calories per capita per day (X5), average consumption of per capita protein per day (X6). While the negative variables used are the percentage of people living below the poverty line (X2). In Cluster 1 most of the variables have a classification of food security in the middle level, except X1 (ratio of food crops availability to normative consumption per capita) and X5 (average consumption of calories per capita per day) which have a high level classification, while X3 (GRDP per capita) have low level classification. This means that the province in the classification of middle food security have good ratio of food crops availability to normative consumption per capita and average consumption of calories per capita per day, but still low in terms of the purchasing power of represented by GRDP per capita which is an approach of per capita income.
In cluster 2 almost all variables have a classification of high food security levels, except X1 (ratio of food crops availability to normative consumption per capita) has a classification of low food security level. So that it can be concluded that the provinces included in cluster 2 are provinces that have high food security classification. Condition ratio of food crops availability to normative consumption per capita that have low level, means that provinces in Indonesia which have high food security have a low food availability. This condition is happen because food availability not only comes from the food corps but it also comes from province's domestic food productions and inter-provincial food imports. However, data on the number of domestic food productions and food imports between provinces is difficult to obtain. While in cluster 3 all variables have a low level of food security classification. So it can be concluded that the provinces included in cluster 3 have a classification of low food security levels.
To improve the occurrence of food security inequality between provinces in Indonesia, the government can focus more on provinces with a classification of low levels of food security in implementing development programs. The variables that need to get more attention are the variables whose conditions are far below the overall average, namely the GRDP Per Capita (X3) variable which is an indicator of food access.
GRDP Per Capita is commonly used as a per capita income approachment that can reflect people's purchasing power, the higher GRDP Per Capita the higher people’s purchasing power the easier people’s food access in the Province. Provinces that fall into the category of middle food security (cluster 1), are Aceh Province, Sumatera Utara Province, Sumatera Barat Province, Sumatera Selatan Province, Bengkulu Province, Lampung Province, Jawa Barat Province, Jawa Tengah Province, Yogyakarta Province, Nusa Tenggara Barat Province, Kalimantan Barat Province, Kalimantan Selatan Province, Sulawesi Utara Province, Sulawesi Tengah Province, Sulawesi Tenggara Province, Gorontalo Province, and Sulawesi Barat Province.
While Provinces that fall into the category of high food security (cluster 2), are Riau Province, Jambi Province, Bangka Beliung Province, Kepulauan Riau Province, Jakarta Province, Banten Province, Bali Province, Kalimantan Tengah Province, Kalimantan Utara and Kalimantan Timur Province. Almost all are provinces that fall into categoriy high food security, have a large potential economic. It can be seen on the value of GRDP of those Provinces that have higer value than other Provinces that fall into cluster 1 nor cluster 3.
Provinces that fall into the category of low food security are Nusa Tenggara Timur Province, Maluku Province, Maluku Utara Province, Papua Barat Province, and Papua Province. Those Provinces low level of all variables of food security that used on this study. This means that those Provinces are reasonable to be priority provinces for immediate implementation of food insecurity prevention policies to improve their achievement of food security and in the long term it will improve indonesian food security state.
Conclusions
The results of the study using based clustering model, produced 3 clusters based on the classification of food security levels. The first cluster consists of 19 provinces with the classification of the level of middle food security, the second cluster consists of 10 provinces with a classification of high food security levels, and the third cluster consists of 5 provinces with a classification of low food security levels. The variable that needs to get more attention is the GRDP Per Capita (X3) which is an indicator of food access. To improve the occurrence of food security inequality between provinces in Indonesia, the government can focus more on provinces with a low level of welfare level in implementing a program to strengthen food security. The program to strengthen food security that is designed and implemented should be more focused on programs to increase food access.
