Prediction Model of Sacha Inchi Crop Development Based on Technology and Farmers’ Perception of Socio-Economic Factors

By inergency On Mar 25, 2024

[ad_1]

2.4. Research Method

This study used a field survey method. Sampling was conducted using non-probability sampling, or more precisely purposive sampling, with the consideration that farmers in both research locations are currently concerned with the development and cultivation of sacha inchi. The sample was given a questionnaire containing a list of questions about the variables that, according to theory, are considered to influence the development of sacha inchi cultivation. In this study, it was first seen how the potential for the development of Sacha Inchi agribusiness in terms of socio-economic aspects. This is done by looking at the perceptions of sample farmers involved in the cultivation of sacha inchi plants. Perceptions are measured using a Likert scale from 1–5, ranging from strongly agree to strongly disagree with the statements in the questionnaire. The questions in this questionnaire were created based on previous research references that were considered relevant to the real conditions in the field. In addition to being given a questionnaire, to support or cross check each farmer’s questionnaire answers, in-depth interviews with farmers and related parties were also conducted. From there it could be illustrated how the socio-economic conditions of farmers could affect the development of sacha inchi plants in the research location. Then, to be able to produce production with high quality and productivity, experimental research was carried out through several fertilization treatments with the aim that researchers could model the optimal fertilization treatment to support the development of massive sacha inchi agribusiness. The collaboration of the two methods is considered very important to do considering the cultural values of farmers in Indonesia who do not easily accept new innovations. Therefore, to be able to develop new commodities massively, it is necessary to deepen the cultural values of farmers in the local area. High product productivity, good quality, and high economic value do not necessarily guarantee farmers to want to develop a particular commodity, including sacha inchi. Meanwhile, the sample of sacha inchi plants that used the treatment of goat manure liquid organic fertilizer and sacha inchi waste liquid organic fertilizer was determined to be 200 trees, which were located randomly in two planting areas. The data collection technique of this research can be seen in the stages of the research flow, as shown in Figure 1.

The initial stage of this research was to identify the problem to be studied. The problem identification was carried out in both regions of West Java Province by focusing on building a model of sacha inchi plant development based on fertilizer treatment and farmers’ perceptions of socio-economic factors. Previously, a preliminary study was conducted to select the method to be used for the creation of the sacha inchi plant development model based on fertilizer treatment and socioeconomic factors, namely, the adaptive neuro-fuzzy inference system (ANFIS) and partial least squares structural equation modeling (PLS-SEM).

Questionnaires to assess the farmers’ perceptions were created in the form of closed questionnaires. The questionnaire contained statements that corresponded to the indicators of each variable of social and economic factors. The questionnaire consisted of approximately 23 questions, of which the social factor variable consisted of 6 questions, the economic factor variable 3 questions, and the sacha inchi agribusiness development variable 14 questions. The questions contained indicators of research variables compiled based on previous research, which were adjusted to the real conditions in the field during the research. Each respondent chose one of the answers from 5 answer options that were provided, ranging from very positive answers to very negative answers. As for interviews with farmers and related parties, the questions were arranged with the development of a questionnaire that was created but did not extend beyond the boundaries that were set. At the time of this interview, the respondents could explain more broadly in response to the questions asked that were not covered in the answer options in the questionnaire.

The data processing began by dividing historical data into two parts: 150 as training data and 50 as checking data. Furthermore, the data were processed using MATLAB R2013a software. For data processing using the adaptive neuro fuzzy inference system (ANFIS) method, the data were first grouped into intervals. The results of the ANFIS and PLS-SEM analyses were then used as the basis for determining the development model of sacha inchi plants based on fertilizer treatment and perceptions of socio-economic factors.

Data Analysis Techniques

The analysis techniques used in this quantitative descriptive research were the adaptive neuro fuzzy inference system (ANFIS) and PLS-SEM methods. The ANFIS method was applied to model the optimal type of fertilizer treatment, while the PLS-SEM method was used to analyze the influence of the farmers’ perceptions of significant socio-economic factors so that from these two analyses, a model of sacha inchi plant development could be built based on fertilizer treatment and perceptions of socio-economic factors.

The Adaptive Neuro-Fuzzy Inference System (ANFIS)

The adaptive neuro fuzzy inference system (ANFIS) is an adaptive neural network based on a fuzzy inference system [31]. ANFIS was applied in this research because it is simple to comprehend, is highly adaptable, tolerates inappropriate data, is capable of modeling nonlinear data, and is able to build and directly apply the expertise of experts [32]. The method of ANFIS offers the benefit of modeling human knowledge’s qualitative side and the decision-making process’s mechanism through given commands [33,34]. Since artificial neural networks are based on incorporated historical data and can predict future events based on those data [35], they have the advantage of being able to recognize specific patterns, learn something unknown, and provide solutions for problems without the requirement for mathematical modeling [36]. ANFIS has the ability to learn by interpretation, which results in a powerful modeling tool [37,38], and automatically generates if–then rules with appropriate membership functions [39]. Figure 2 below shows the layer structure of ANFIS.

Figure 2 shows that there are 5 layers, where layer 1 is input data consisting of fertilizer use, the number of leaves, the number of stems, and potential fruit. In the second layer is the model formation by ANFIS, in layer 3 is the data testing process by the ANFIS application, in layer 4 is prediction testing based on training data and real data that have been entered into the application, and layer 5 is the output of the ANFIS prediction model results.

Figure 2 also shows there are two types of nodes: adaptive nodes with square icons and nodes with circle icons. The output of each layer is denoted by Oj, where O is the number of rules, and j is the number of layers.

The ANFIS network consists of five layers as follows [25,40,41].

Layer 1 (fuzzy layer)

Each node in layer 1 is an adaptive node, which means that the parameter value can change with the following node function:

O1, 1t = μA1 (Zt − 1)

(1)

O1, 2t = μA2 (Zt − 1)

(2)

O1, 3t = μB1 (Zt − 2)

(3)

O1, 4t = μB2 (Zt − 2)

(4)

where (Zt − 1) and (Zt − 2) are the inputs at the i-th node. Meanwhile, −1 and −2 are the membership functions of each node. The degree of membership of each input to fuzzy sets A and B is expressed as 1, with 1, 2, 1, 2 being linguistic variables. The membership function used is the generalized bell membership function. The generalized bell membership function can be written as follows:

$μ_{A} (x_{t}) = e^{\frac{{- (x_{t} - c)}^{2}}{{2 σ}^{2}}}$

(5)

$f x; a_{i}, b_{i}, c_{i} = \frac{1}{1 + \frac{{x - c_{i}}^{{2 b}_{i}}}{a_{i}}}$

(6)

in which AND are a set of parameters called premise parameters. By taking the value = 1, only the parameters AND will change during the process of learning. The change in the values of these parameters will also change the generalized bell curve.

Layer 2 (product layer)

Each node in layer 2 is a non-adaptive node, which means the parameter values are fixed. The function of this node multiplies each incoming input signal as follows:

$O_{2, i} = w_{i} = μ A_{i} (x) \cdot μ B_{i} (y); i = 1, 2$

(7)

Each output node expresses the activation degree of each fuzzy rule. The number of rules formed follows the number of nodes in this layer.

Layer 3 (normalization layer)

Each node in this layer is a non-adaptive node that states a normalized degree function that is the ratio of the i-th node output in the previous layer, which is written as follows:

$O_{3,1} = w_{t} \frac{w_{i}}{w_{1} + w_{2}}, a n d i = 1, 2$

(8)

The function can be expanded if there are more than two rules by dividing it by the total number of w for all rules.

Layer 4 (defuzzification layer)

Each node in this layer is an adaptive node with the following node function:

$O_{4,1 t} = w_{1 t}^{*} Z_{t}^{(1)} = w_{1 t}^{*} (α_{1} Z_{t - 1} + β_{1} Z_{t - 2} + γ_{1}),$

(9)

$O_{4,2 t} = w_{2 t}^{*} Z_{t}^{(2)} = w_{2}^{*} (α_{2} Z_{t - 1} + β_{2} Z_{t - 2} + γ_{2})$

(10)

where $α_{i}$ , $β_{i}$ , $γ_{i}$ is the set of parameters of the node and is called the consequent parameter.

Layer 5 (total output layer)

Layer 5 is the last layer that functions to add up all inputs with the following node function:

$O_{5 t} = {\hat{Z}}_{t} = w_{1 t}^{*} Z_{t}^{(1)} + w_{2 t}^{*} Z_{t}^{(2)}$

(11)

Partial Least Squares Structural Equation Modeling (PLS-SEM)

This study used the PLS-SEM analysis technique to analyze the effect of the perception of socio-economic factors on the development of sacha inchi plants. The significance level used for this research was 95%. The statistical measurement scale used in this research was an ordinal measurement scale. This study used the ANFIS and PLS-SEM approaches since the two approaches reinforced each other; when the SEM results only measured how great the influence was, it was strengthened by the results of the ANFIS approach.

[ad_2]