1. Introduction
India is an agronomy-based nation, and more than 50% of the country’s employment is accomplished through agriculture [
1]. According to the worldometers report [
2], in 2019, the Indian population is reported to be 1.37 billion, and it is increasing at a prominent pace. Further, this, in turn, will demand the farmers to generate a higher amount of cultivable produce. Also, the primary point to be noted here is that only 60 percent of the land is suitable for doing cultivation. Based on the prediction by the econometric models [
3] the gross domestic product (GDP) of India’s agriculture by 2020 is expected to reach approximately INR 7000 Billion. It is clearly evident from these figures that India’s significant chunk of GDP is dependent on agriculture.
Furthermore, farmers in India come across numerous problems and issues such as frequently varying weather, typhoons, floods, tsunami, landslides, earthquakes, and poor quality of the soil. Therefore, to overcome such challenges, in addition to traditional knowledge, the farmers might require support from various emerging technologies and decision systems. Moreover, the failure of agriculture might also influence the production of allied businesses, and all these factors put together will have a strong influence on the nation’s GDP and growth. Also, it is essential to keep the inflation in check by continually monitoring and supporting the different contributing areas of the nation’s GDP. It is a fact that in third world nations’ such as India, the agriculturists lack quality education and cognizance about the modern-day technological advancements and practices in farming. Further, it becomes essential for the agriculturists to make several decisions such as choosing the type of crop, based on numerous factors like seasons, rainfall, soil quality and condition, and so on. Hence, technology-supported decision-making becomes a significant part of the farming process. Several decision-making models have been developed so far in the field of agriculture [
4,
5,
6,
7,
8].
The sharp increase in population is much more severe in some countries around the world, particularly in India. There is a demand for food being increased every year along with the increase in population. Further, this can be solved by applying sustainable agriculture practices. The selection of land for crop cultivation has a significant impact on sustainable agriculture. An essential requirement for long term productivity and profitability obtained from farming in rural areas is sustainable agriculture development. Sustainable agriculture includes social, financial, and environmental aspects. Crop productivity is one of the essential elements in the financial aspects of farming [
9]. Identification of suitable farm for crop cultivation is essential to maximize the productivity of crops. Crop production relies on multiple criteria that may differ from place to place. The present work emphasis on the sugarcane farm selection that considers comprehensive parameters for sustainable farming.
Several decision models have been developed so far for the development of sustainable agriculture. A decision support model was developed for the selection of cropping pattern using Fuzzy and multi-criteria decision making (MCDM) approaches for sustainable agriculture development [
9]. A stochastic decision model was developed using multi-criteria decision analysis to select sustainable biomass crop for the production of biofuels with multiple conflicting criteria [
10]. A decision model was proposed to rank sustainable energy conversion technologies, which convert agriculture residues to energy using Fuzzy analytic network process (ANP) and VIKOR (Vlsekriterijumska Optimizacija I Kompromisno Resenje) methods [
11]. A model was developed to identify the best water supply management alternative for sustainable agriculture using VIKOR, and Fuzzy Order weighted average methods for climate-change adaptation [
12]. A hybrid model was developed to select sustainable supply chain for Agri produce in India using Interpretive Structural Modeling (ISM), decision making trial and evaluation laboratory (DEMATEL), analytic network process (ANP) methods [
13].
Multi-criteria decision making (MCDM) approaches are used to develop useful decision-making tools when multiple conflicting criterions are taken into consideration for a given problem. In the existing literature, several MCDM methods have been used for ranking and predicting the best solution among the specified alternatives. A Technique for Order Preference by Similarity to an Ideal Solution (TOPSIS) is a renowned MCDM technique with the capability of choosing the best alternative in close proximity with the positive ideal solution and at the same time far away from the negative ideal solution [
14,
15]. The Vlsekriterijumska Optimizacija I Kompromisno Resenje (VIKOR) is a popular and pragmatic MCDM technique, which can be employed for resolving decision problems in business management, sustainable agriculture, supply chain networks, production and design, and so on [
16,
17]. Further, the grey relational analysis (GRA) is a well-known MCDM technique that amalgamates together the quantitative and qualitative information which assumes either the largest or smallest evaluation criteria value [
18,
19]. A simple additive weighting (SAW) is a straightforward technique in which the preference measures are determined for every alternative using the sum of products of weights and alternative values [
20,
21].
Several decision models have been developed by integrating MCDM methods. A decision model was developed using 11 parameters to predict green supplier in the supply chain using VIKOR and TOPSIS [
22]. VIKOR and TOPSIS methods have been used to develop a ranking model to rank and monitor the countries based on their performance [
23]. An integrated MCDM model was developed using AHP [
24] and VIKOR for the selection of Hazardous Waste Carrier to provide a clean environment and to reduce the transportation cost [
25]. A group decision-making method was developed by integrating ELimination and Choice Expressing REality(ELECTRE) [
26] and VIKOR method for solving outranking problems with uncertainty [
27]. AHP and TOPSIS have been integrated with Fuzzy in order to assess the Human Resource in Asian countries [
28]. In this VTOPES (VIKOR, TOPSIS, entropy, standard deviation) model, VIKOR and TOPSIS approaches are integrated for decision making.
In general, feature selection is essential in decision making, where classification and predictions are involved [
29]. In MCDM, weight computation plays a vital role in the decision-making process. When multiple parameters are involved in decision making, it cannot be concluded that each parameter has the same meaning and importance. Therefore, it is mandatory to compute the weights of parameters in the decision-making process. In practice, there are numerous approaches for determining the weights of various parameters. In general, the determination of the weights can be classified into two major types, the first is said to be the subjective approach, and the second one is termed as the objective approach. The computation of the weights in the first approach is based on specialists’ advice, substantiation, and validation. Whereas, in the second approach, the computation of weights is accomplished with the support of mathematical models.
Shannon’s entropy approach is an objective weight calculation approach usually deployed to handle uncertain problems [
30]. The correlation coefficient and standard deviation (CCSD) technique is a weight computation approach in which the criterions’ objective weights are calculated [
31]. Standard deviation (SD) is another simple objective weight calculation method which considers mean and standard deviation of parameters for weight assignment. Analytical hierarchy process (AHP) is one of the MCDM approaches and is also used for subjective weight calculation of parameters [
32]. Weighted least square method is another popular subjective method used for weight calculation of parameters [
33]. Another subjective assignment method which assigns weight by considering the experts’ opinion is the Delphi method [
34,
35]. In this work, Shannon’s entropy method and SD method are used for the weight assignment of parameters.
The key contributions of this work are summarized as follows:
- (i)
This proposed VTOPES model is a general approach that aims to predict the best alternative in decision-making problems effectively and can be further applied to other decision problems.
- (ii)
In this work, the significant scores of each parameter are evaluated based on the proportional weights for better decision making.
- (iii)
Even though sub-parameters have a strong influence on the decision-making process, they are seldom considered in many of the previous research works. Hence, it is necessary to have a hybrid model that will consider both the main parameter and sub-parameters in order to select the best sustainable sugarcane farms.
- (iv)
As of now, there is no generic multi-criteria decision-making model for solving decision problems characterized by multiple conflicting criteria.
- (v)
The proposed work offers superior performance in terms of ranking results when compared with the other existing MCDM methods.
The organization of the paper is as follows: In the next section, the steps involved in the development of VTOPES model is explained. In
Section 3, the results of VTOPES model has been discussed and further validated with other MCDM methods.
2. Proposed VTOPES Model
The VTOPES decision model is proposed to predict the best alternative from the given set of alternatives which are characterized by multiple parameters. The VTOPES model considers the significance score of the parameters with respect to the alternatives and takes the appropriate decision. Entropy and standard deviation methods have been applied to find the significance scores of the parameters. VIKOR and TOPSIS methods are applied to generate final assessment values of alternatives which are used to rank the given set of alternatives. The list of symbols and notations used in the VTOPES model is depicted in
Table 1.
The raw data obtained for decision making consists of m alternatives and multiple parameters. Multiple parameters identified for the decision-making problem are grouped into g main parameters, and each main parameter has its own sub-parameters. Each main parameter matrix Y consists of m alternatives and n sub-parameters.
The proposed VTOPES model is described as follows:
VTOPES model
For g main parameters matrix Y with m alternatives and n sub-parameters do the following:
Determination of Entropy
where
is the entropy constant and defined as
,
n is the number of sub-parameters and
m is the number of alternatives.
Calculation of Degree of Divergence
where
D is the degree of divergence.
Sub-parameter weight calculation
The weight of
sub-parameter,
, is defined as
Here .
The measure of utility computation
where
and
A is utility measure
The measure of Regret Computation
The regret measure is obtained by
Generation of Ranking Indices of Main parameters denotes the ranking index of alternative with respect to the main parameter.
is strategic weight representing maximal group utility, and is the weight of each regret measure.
;
Determination of Ranking Indices Matrices
The Ranking Indices Matrix with
m alternatives and
g columns is represented as.
where
Ct is partitioned columnar sub-matrices of
X initialized with value zero and
.
Standard Deviation Computation
where
and σ represents mean and standard deviation, respectively.
Calculation of main parameter weights
The weight of
jth main parameter is calculated by.
where
and
.
Computation of weighted normalized decision matrix
The weighted normalized decision matrix is calculated by
where
and
.
Calculation of Positive Ideal solution
The positive ideal solution (most preferable alternative) is obtained by
Calculation of Negative Ideal Solution
The negative ideal solution (least preferable alternative) is obtained by
Determination of Separation Measure
The separation distance of each alternative from the positive ideal solution is determined by
The separation distance of each alternative from the negative ideal solution is determined by
Computation of Final Assessment Values
The final assessment values of the given alternatives can be obtained by
where
and
.
The proposed system model is applied to agriculture dataset for ranking the sustainable sugar cane farms in terms of yield. The sugarcane farm dataset is obtained from arbitrarily identified villages located east of Eastern Ghats in Tamil Nadu, India. Primarily, 27 parameters have been identified for sustainable farming based on the recommendations made by agricultural experts. The 27 parameters have been grouped into seven main parameters, each having its own sub-parameters, as mentioned in
Table 2 with notations used in this work. The associative significance of various sub-parameters for the respective main parameter occupies a vital part of the decision-making process. In this model, distinct approaches are utilized for computing the sub-parameters and main parameter’s weights.
Generally, the source data gathered from the identified agricultural sites have to be segregated into sub-parameter matrices for every main parameter. It has to be noted that the source data gathered from farms have distinct units of measure. Therefore, for normalizing the source data, and also for ranking the identified sustainable sugarcane farms based on the main parameter, the ranking indices of all the main parameters are calculated. The VIKOR technique is utilized for determining the main parameters measures of ranking indices matrix. Hence, it can be observed that the source data available in the sub-parameter matrix is transformed into the main parameter measures of preference matrix.
Subsequently, the standard deviation technique is deployed for calculating the main parameters’ associative weights. Furthermore, the main parameters’ weights and its measure of ranking indices matrix are then employed to the TOPSIS approach for determining the assessment values of the identified sustainable sugarcane farms. Consequently, the farms are ranked by using the obtained assessment values.
The proposed ensemble VTOPES model is segregated into five different phases as follows:
2.1. Phase 1
Grouping the parameters into two different groups, namely the main parameters and sub-parameters (
Table 2).
2.2. Phase 2
Computing the associative significance of sub-parameters for every corresponding main parameter. Shannon Entropy method is used for this purpose. In this objective weight assignment method, the entropy concept is used to measure the uncertainty occur in data. If the entropy value of a parameter is higher, then its corresponding weight will be smaller. Thus, the weights of sub-parameters are calculated using the entropy method and shown in
Table 3.
As the raw data obtained is in the form of sub-parameter matrix, it is desirable to combine the sub-parameter values into main parameter scores.
2.3. Phase 3: Determining the Main Parameters’ Ranking Indices of Identified Agricultural Farms
Since the raw data may contain conflicting parameters, it is desirable to find compromise ranking solution. VIKOR is popular MCDM method used for finding compromise solution by considering maximum utility scores and minimum individual regret scores of the given alternatives. The weights of sub-parameters under each main parameter (
Table 3) and their corresponding sub-parameter matrix are considered for finding the main parameters’ ranking indices of sustainable sugarcane farms. The compromise solution is obtained by comparing the closeness measure of the optimal alternative [
16]. The utility scores, regret scores, and ranking indices obtained for soil main parameter are shown in
Table 4. Thus, the ranking indices of each main parameter is obtained to form the final ranking indices matrix, which is shown in
Table 5.
2.4. Phase 4: Main Parameters’ Weight Computation
A simple weight computation method, namely the standard deviation (SD) method, is used in this model for the computation of main parameter weights and shown in
Table 6.
2.5. Phase 5: Determining the Assessment Values of Identified Agricultural Farms
The final assessment values need to be determined in order to evaluate and rank the identified 20 sustainable sugarcane farms. There are possibilities that real-world data may contain imprecise and incomplete information. TOPSIS is popular MCDM method used to handle such data and provide reliable decisions. The main parameter weights obtained in phase 4 (
Table 6) and ranking indices matrix obtained in phase 3 (
Table 5) are applied to TOPSIS initially to compute weighted normalized matrix shown in
Table 7. It considers the positive ideal solution or most preferable alternatives and negative ideal solution or least preferable alternatives (
Table 8) for the evaluation of alternatives (farms). Then it measures the separation distance of each alternative from a positive ideal solution and negative ideal solution shown in
Table 9. This distance measure is used to find the proximity measure or assessment values of alternatives (
Table 9).
3. Results and Discussions
The ranking results obtained using the assessment values calculated using VTOPES model is shown in
Table 10. Also, these ranking outcomes obtained from the proposed system model are compared with the results of widespread MCDM techniques such as the grey relational analysis (GRA) and simple additive weight (SAW) approaches. Further, the previous five years mean yield/hectare data gathered from the sustainable sugarcane farms are utilized for validating these ranking outcomes.
The results thus obtained from VTOPES model, the GRA and SAW are compared with the ranking pattern acquired from the sustainable average yield data and represented in graphical format (
Figure 1). In this graph, F1, F2, F3, etc., denote the identified agriculture farms as given in
Table 10. Further, it can be noticed that the ranking patterns obtained from the average yield are in the form of a straight line. However, the ranking results of the VTOPES model shows minor deviations from the straight line. Besides, the GRA and SAW ranks depict significant deviations from the straight line in comparison with the VTOPES ranks.
Moreover, Spearman’s rank correlation coefficient method is employed to test the significance of correlation among the ranks obtained from the developed model and the yield data. The value ⍴ calculated for the developed model VTOPES model is 0.995, GRA is 0.975, and SAW is 0.961. The critical value ⍴ for the datasets with 20 trials with α value 0.01 is 0.544. Therefore, the null hypothesis is rejected because the ⍴ value acquired from the Spearman’s rank correlation method is higher than the critical value ⍴. Thus, the correlation between the ranking pattern acquired using the developed VTOPES model and the ranks obtained from the yield data is considered to be significant with 99% confidence level.
The VTOPES model correctly ranked 17 farms out of 20 agricultural farms, thus obtaining 85% of accuracy (
Table 10). Whereas predictive mathematical model [
6] accurately predicted 16 farms out of 20 farms and obtained 80% of accuracy. With respect to the time, as a predictive mathematical model integrates three weight calculation methods, two MCDM methods, it took more time to obtain decision results when compared to the proposed VTOPES model. Hence, the developed VTOPES model proves to produce accurate results for the given sustainable sugarcane farms, and it can be further applied to solve any MCDM problems where multiple parameters are considered.