Received: 03 Jun 2025 Revised: 16 Dec 2025 https://doi.
org/10.
13057/ijas.
Accepted: 20 Dec 2025
Available online: 27 Feb 2026
E-ISSN: 2621-086X
https://jurnal.
id/ijas Indonesian Journal of Applied Statistics Vol.
No.
2, pp.
113-126, 2025
Vol 8.
No 1, 1-10, 2025 Comparative Analysis of Fuzzy Mamdani Method and Fuzzy Sugeno Method in Predicting Household Electricity Consumption Costs Luthfia Zahra*.
Mashuri Department of Mathematics.
Universitas Negeri Semarang.
Semarang.
Indonesia *Corresponding author: luthfiazahra8@gmail.
Abstract Electricity has become an essential part of our daily lives.
As technology has rapidly developed, many modern activities and devices have become highly dependent on electricity.
The more electricity that is used, the higher the monthly cost.
This cost is influenced by usage patterns and various uncertain factors.
Fuzzy logic is one approach that can be used in decision support systems in the face of uncertainty like this.
This study aims to apply the Mamdani and Sugeno fuzzy methods based on house building area, number of electronic devices, number of family members, and income to determine which method more accurately predicts household electricity consumption costs based on the mean absolute percentage error (MAPE) value.
Data for this study were obtained through questionnaires and interviews with residents of Margorejo Village.
Data processing yielded a MAPE value of 12.
3% for the Mamdani method and a MAPE value of 9.
9% for the Sugeno method.
Based on these results, the MAPE value for the Sugeno method is smaller than that for the Mamdani method.
Therefore, it can be concluded that the Sugeno method is more accurate for predicting household electricity consumption costs in Margorejo Village.
Keywords: Mamdani method.
Sugeno method.
MAPE.
This is an open access article under the Creative Commons Attrribution-ShareAlike 4.
0 International License How to Cite:
Zahra and Mashuri.
AuComparative analysis of fuzzy Mamdani method and fuzzy Sugeno method in predicting household electricity consumption costs,Ay Indonesian Journal of Applied Statistics, vol.
8, no.
2, pp.
113-126, 2025, doi: 10.
13057/ijas.
INTRODUCTION
Electricity is a major necessity in people's lives, used by businesses, institutions, and the general public to carry out various activities.
Almost all equipment, from households to transportation, depends on electrical energy, so its availability is very important every day .
Electricity demand increases with population and economic growth, affecting the amount of electricity that must be produced to meet demand .
Kabupaten Pati, located in the eastern part of Central Java, has an area of 150,368 Based on topographic data, most of the area is in the lowlands with an altitude of 0-100 meters above sea level, covering approximately 100,769 hectares .
Pati Regency is experiencing rapid economic growth that has a positive impact on the industrial and electricity sectors.
PT PLN UP3 Kudus recorded a 5.
76% increase in electricity sales in this region, a figure that is relatively high compared to other districts in Central Java.
The cost of electricity consumption is influenced by several factors, such as house size, electrical power, number of electronic devices, family members, and income level.
By understanding these factors, a mathematical approach can be used to estimate monthly electricity costs more accurately.
Mathematics is a method of logical thinking with various branches, such as statistics, algebra, and One important concept in mathematical logic is fuzzy logic, which is used to address ambiguity and uncertainty in real systems.
This concept is very relevant in everyday life because many problems involve uncertainty .
Household electricity cost prediction is a fuzzy problem because it involves uncertainty and nonlinear relationships between various factors, such as house size, number of electronic devices.
Copyright A 2025.
Author.
E-ISSN: 2621-086X https://doi.
org/10.
13057/ijas.
IJAS: Indonesian Journal of Applied Statistics.
Vol.
No.
Luthfia Zahra.
Mashuri number of family members, and income.
For example, a large house generally requires more electricity, but a small house may consume more electricity depending on its efficiency.
Differences in power and usage patterns of electronic devices, electricity needs of each family member, and lifestyle based on income also add to the complexity.
Each factor affects electricity consumption .
Fuzzy logic, which is based on fuzzy set theory, has various applications in everyday life such as prediction of scholarship candidates, optimization of drug demand, and scheduling of lectures.
One of its applications is as a decision support system to predict electricity usage costs.
In its calculation, fuzzy logic can use several methods, including the Mamdani.
Sugeno, and Tsukamoto methods, each of which has a different approach and calculation results.
tion differently, creating uncertainty in the calculation .
The Mamdani fuzzy method, introduced by Ebrahim Mamdani in 1975, is popular for its ability to handle uncertainty and is easy to understand.
This method consists of four main processes to produce the output and is often completed using Matlab software, which provides various tools for the Mamdani Meanwhile, the Sugeno fuzzy method, introduced by Takagi-Sugeno Kang in 1985, has many similarities with Mamdani but differs in the form of output.
The Sugeno method produces output in the form of constants or linear equations, while Mamdani produces output in the form of fuzzy sets.
Matlab also provides special tools for the application of the Sugeno method.
There have been many studies conducted related to efforts to predict electricity usage, including by Aenun .
Santosa .
, and Haryanto .
In addition, research using the Mamdani method fuzzy logic has been carried out by Yudhistira et al.
, .
Tundo & Mahyuzar .
Nurhayati et al.
, .
Research using fuzzy logic Sugeno method has also been widely done, including by Pasaribu & Rozy .
Simanjuntak .
Hafiz & Sriani .
Based on research conducted by Aenun and Mashuri .
using the Mamdani fuzzy method with 4 input variables, namely house area, electrical power, electronic devices, and economic income, and the help of Matlab software, the results show that the Mamdani method is well used to predict electricity usage costs.
Then, based on Santosa's research .
using the Sugeno fuzzy method with 1 input variable, namely power usage, the results show that the Sugeno fuzzy method allows the calculation of electricity tariffs with a better level of smoothness and provides results that are considered fairer by consumers.
Based on the description above, the researcher intends to conduct research on the comparison between the Mamdani fuzzy method and the Sugeno fuzzy method in predicting electricity consumption costs with the title AuComparative Analysis of the Mamdani Fuzzy Method and the Sugeno Fuzzy Method in Predicting Household Electricity Consumption Costs.
METHODS
This research method includes literature study, problem formulation and solving, and conclusion Prior to problem solving, data was collected through questionnaires and interviews in RT 002.
Margorejo Village.
The number of samples taken for this study was 50 respondents, where 10 respondents were obtained through interviews and 40 respondents were obtained through The respondents interviewed came from communities with different occupational segments, and therefore had varying levels of income.
In addition, the respondents also had varying household characteristics, including house size, number of electronic devices, and number of family members, which were relevant to the research objectives.
The data collected included building area .
, number of electronic devices .
, number of family members .
, income (Rupia.
, and monthly electricity consumption costs (Rupia.
This research uses four variables input, namely the input of building area, number of electronic devices, number of family members, and income and the output of electricity consumption costs.
The cost prediction is done using Mamdani and Sugeno fuzzy methods with the help of Matlab R2014a Copyright A 2025.
Author.
E-ISSN: 2621-086X https://doi.
org/10.
13057/ijas.
IJAS: Indonesian Journal of Applied Statistics.
Vol.
No.
Luthfia Zahra.
Mashuri Mamdani Method In this research, the method applied is the Mamdani method, which was introduced by Ebrahim Mamdani in 1975 and is also known as the Max-Min method.
To obtain the output of this method, there are 4 stages of the process that must be completed, namely:
Fuzzification is the process of transforming input from a firm set .
into a fuzzy set .
This stage determines the fuzzy set of each input and output variable using a membership Implication Function Application.
In the Mamdani method, the implication function used is min.
The basic rules formed are 81 rules based on each input variable.
Rule Composition, in the Mamdani fuzzy method the inference process is carried out using the max method.
Defuzzification, in the Mamdani method using the centroid method, with the following formula O yyA.
yeu yeIyeu yeuO = yeu Oyeu yyA.
yeIyeu where yc is represents the i-th domain value, yuN.
is indicates the degree of membership at that point and yc O is the result of defuzzification.
Sugeno Method This research uses the zero-order Sugeno method with the help of Matlab R2014a software with the following stages:
Fuzzification.
This stage determines the fuzzy set of each input and output variable using the membership function.
Implication function application.
Fuzzy rules in the Sugeno fuzzy method are the same as the Mamdani method which is formed by 81 rules.
Rule Composition, in the Sugeno fuzzy method the inference process is carried out using the max Defuzzification, the Sugeno method uses the weighted average method.
After the calculations are carried out using the Mamdani and Sugeno fuzzy methods manually and with the help of Matlab R2014a software, the next step is to calculate the mean absolute percentage error (MAPE) value.
The use of mean absolute percentage error (MAPE) in evaluating forecasting results can see the level of accuracy of forecasting figures and realization figures .
The MAPE value can be calculated using the following formula:
ycU Oe yaycn Ocycuycn=1 | ycn ycUycn | .
ycAyaycEya = y 100% ycu ycUycn = the original data value of the ycn-th observation yaycn = the forecast value of the ycn-th observation ycu= the amount of data The accuracy level of the forecasting results can be categorized based on the MAPE value which can be seen in more detail in the following Table 1.
Table 1.
MAPE value evaluation MAPE (%) <10 10 Ae 20
20 Ae 50
>50
Copyright A 2025.
Author.
E-ISSN: 2621-086X Interpretation Very accurate prediction Good prediction Reasonable prediction Inaccurate prediction https://doi.
org/10.
13057/ijas.
IJAS: Indonesian Journal of Applied Statistics.
Vol.
No.
Luthfia Zahra.
Mashuri The final step of the research is to draw conclusions based on the lowest MAPE value to determine the most accurate method.
RESULTS AND DISCUSSION
In this study, the input variables used are the size of the house, the number of electronic devices, the number of family members, and income, with the output being the cost of electricity consumption.
Fuzzy sets have a domain of values in the universe of discourse.
Generally, this domain is determined using an expert system, which is a system that stores and applies expert knowledge to provide However, in this study, the determination of the domain of each fuzzy set was carried out directly by the researcher .
The fuzzy set of each variable is formed based on the data obtained as shown in Table 2.
Table 2.
Fuzzy set of research variable Function
Input
Output
Variable Name
House Building Area
Fuzzy Set
STANDARD
MEDIUM
LARGE
Number of FEW
Electronic NORMAL
Devices MANY Number of FEW
Family NORMAL
MANY
Income
LOW
MEDIUM
HIGH
Electricity Cost
LOW
MEDIUM
HIGH
Universe of Discourse .
, .
, .
, .
, .
, 1.
Domain , .
, .
, .
, .
, .
, .
, .
, .
, .
, .
, .
, .
, .
, .
, 1.
Application of Mamdani Fuzzy Method The calculation will use data from the 14th respondent with a house building area of 104 yco2, the number of electronic devices used is 12 units, with a family of 2 people, and a monthly income of Rp9,000,000.
Fuzzification The variable Area of House Building, with a universe of speech .
, .
is divided into three fuzzy sets, namely STANDARD.
MEDIUM, and LARGE.
Membership functions are represented with linear and triangular curves .
resented in Figure .
and are formulated as follows:
Copyright A 2025.
Author.
E-ISSN: 2621-086X https://doi.
org/10.
13057/ijas.
IJAS: Indonesian Journal of Applied Statistics.
Vol.
No.
Luthfia Zahra.
Mashuri Figure 1.
Membership function of house building area 1, yca O 42 100 Oe yca yuNycoycaycycIycNyaycAyayaycIya .
= { , 42 O yca O 100 100 Oe 42 yca Ou 100 0, yca O 80 ycuyc yca Ou 180 yca Oe 80 , 80 O yca O 130 yuNycoycaycycAyayayaycOycA .
= 130 Oe 80 180 Oe yca .
Oe 130 , 130 O yca O 180 0, yca O 150 yca Oe 150 yuNycoycaycyayaycIyaya .
= { , 150 O yca O 330 330 Oe 150 yca Ou 330 Because the house building area of 104 yco is located in the MEDIUM category, the following membership values are obtained:
yca Oe 80 104 Oe 80 24 yuNycAyayayaycOycA .
= = 0.
130 Oe 80 The variable Number of Electronic Devices, with a universe of speech .
, .
is divided into three fuzzy sets namely FEW.
NORMAL, and MANY.
The membership functions are represented with linear and triangular curves .
resented in Figure .
and are formulated as Figure 2.
Membership function number of electronic devices 1, yca O 1 8Oeyca yuNycycyyceyayaycO .
= { ,1 O yca O 8 8Oe1 0, yca Ou 8 Copyright A 2025.
Author.
E-ISSN: 2621-086X https://doi.
org/10.
13057/ijas.
IJAS: Indonesian Journal of Applied Statistics.
Vol.
No.
Luthfia Zahra.
Mashuri 0, yca O 6 ycaycycayc yca Ou 15 ycaOe6 , 6 O yca O 11 .
yuNycycyyceycAycCycIycAyaya 11 Oe 6 15 Oe yca .
Oe 11 , 11 O yca O 15 0, yca O 13 yca Oe 13 yuNycycyyceycAyaycAycU .
= { , 13 O yca O 21 21 Oe 13 1, yca Ou 21 Because the number of electronic devices owned is 12 units, it is located in the NORMAL category, so the membership value is obtained as follows:
15 Oe yca 15 Oe 12 3 yuNycAycCycIycAyaya .
= = = 0.
15 Oe 11 The variable Number of Family Members, with a universe of speech .
, .
is divided into three fuzzy sets namely LITTLE.
NORMAL, and LOTS.
Membership functions are represented with linear and triangular curves .
resented in Figure .
and are formulated as follows:
Figure 3.
Membership function number of family 1, yca O 2 4Oeyca yuNycycaycoyayaycO .
= { ,2 O yca O 4 4Oe2 0, yca Ou 4 0, yca O 3 ycuyc yca Ou 6 ycaOe3 ,3 O yca O 4 yuNycycaycoycAycCycIycAyaya .
= 4 Oe 3 6Oeyca .
Oe 4 , 4 O yca O 6 0, yca O 5 ycaOe5 yuNycycaycoycAyaycAycU .
= { ,5 O yca O 8 8Oe5 1, yca Ou 8 Because the number of family members is 2 people, it is located in the FEW category, so the membership value is obtained as follows:
yuNyayaycO .
= Copyright A 2025.
Author.
E-ISSN: 2621-086X
4Oeyca 4Oe2 2
= =1
4Oe2 https://doi.
org/10.
13057/ijas.
IJAS: Indonesian Journal of Applied Statistics.
Vol.
No.
Luthfia Zahra.
Mashuri The income variable with the universe of speech .
, .
is divided into three fuzzy sets, namely LOW.
MEDIUM, and HIGH.
The following is the membership function for the income variable which is formulated as follows and presented in Figure 4.
Figure 4.
Membership function income 1, ycc O 2 5Oeycc yuNycyyaycCycO .
= { ,2 O ycc O 5 5Oe2 0, ycc Ou 5 0, ycc O 4 ycuyc ycc Ou 8 yccOe4 ,4 O ycc O 6 yuNycyycAyayayaycOycA .
= 6 Oe 4 8Oeycc .
Oe 6 , 6 O ycc O 8 0, ycc O 7 yccOe7 yuNycyyayayaya .
= { , 7 O ycc O 10 10 Oe 7 1, ycc Ou 10 Because his monthly income is Rp9,000,000, it is located in the HIGH category, so the membership value is obtained as follows:
yuNyayayaya .
= yccOe7 9Oe7 2 = = 0.
10 Oe 7 Electricity consumption cost variable In the electricity consumption cost variable with the universe of speech .
, 1.
, it is divided into three fuzzy sets, namely LOW.
MEDIUM, and HIGH.
The following is the membership function for the variable cost of electricity consumption which is formulated as follows and presented in Figure 5.
Figure 5.
Membership function of electricity consumption cost Copyright A 2025.
Author.
E-ISSN: 2621-086X https://doi.
org/10.
13057/ijas.
IJAS: Indonesian Journal of Applied Statistics.
Vol.
No.
Luthfia Zahra.
Mashuri 1, yce O 40 300 Oe yce yuNycaycoycoyaycCycO .
= { , 40 O yce O 300 , 300 Oe 40 0, yce Ou 300 0, yce O 250 ycuyc yce Ou 700 yce Oe 250 , 250 O yce O 475 yuNycaycoycoycAyayayaycOycA .
= 475 Oe 250 700 Oe yce .
Oe 475 , 475 O yce O 700 0, yce O 600 yce Oe 600 yuNycaycoycoyayayaya .
= { , 600 O yce O 1000 1000 Oe 600 1, yce Ou 1000 .
Application of Implication Function Based on the fuzzy set of each input variable, 81 fuzzy rules are formed, as shown in Table 3.
Table 3.
Rules fuzzy Rules Fuzzy House Building Area [R.
[R.
[R.
[R.
[R.
[R.
[R.
[R.
[R.
STANDARD
STANDARD
STANDARD
STANDARD
STANDARD
STANDARD
STANDARD
STANDARD
STANDARD
Number of Electronic Devices FEW
FEW
FEW
FEW
FEW
FEW
FEW
FEW
FEW
Number of Family Income
Electricity Cost
FEW
FEW
FEW
NORMAL
NORMAL
NORMAL
MANY
MANY
MANY
LOW
MEDIUM
TINGGI
LOW
MEDIUM
HIGH
LOW
MEDIUM
HIGH
LOW
LOW
LOW
LOW
LOW
MEDIUM
LOW
MEDIUM
MEDIUM
Based on all the rules that have been established, there is one matching rule:
[R.
If the house area is MEDIUM and the number of electronic devices is NORMAL and the number of family members is FEW and the income is HIGH, then the electricity consumption cost is MEDIUM.
yuNycI.
= yuNycoycaycycAyayayaycOycA .
O yuNycycyyceycAycCycIycAyaya .
O yuNycycaycoyayaycO .
O yuNycyyayayaya .
= min .
uNycoycaycycAyayayaycOycA .
, yuNycycyyceycAycCycIycAyaya .
, yuNycycaycoyayaycO .
, yuNycyyayayaya .
) = min.
= 0.
Rule Composition At the rule composition stage using the MAX method to combine the results of rule composition.
Next, determine the cut-off point of the rule when yuNycaycoycoycAyayayaycOycA = 0.
48 as follows:
Intersection point 1 Intersection point 2 yc1 Oe250 = 0.
475Oe250 700Oeyc2 = 0.
700Oe475 Ni yc1 Oe 250 = 0.
Ni yc1 Oe 250 = 108 Ni yc1 = 108 250 Ni yc1 = 358 Ni 700 Oe yc2 = 0.
Ni 700 Oe yc2 = 108 Ni 700 Oe 108 = yc2 Ni 592 = yc2 Copyright A 2025.
Author.
E-ISSN: 2621-086X https://doi.
org/10.
13057/ijas.
IJAS: Indonesian Journal of Applied Statistics.
Vol.
No.
Luthfia Zahra.
Mashuri Figure 6.
The result of rule composition So that the new fuzzy solution region is obtained as shown in Figure 6 with its new membership function as follows:
yc Oe 250 250 O yc O 358 475 Oe 250 yuNycaycoyco .
= 358 O yc O 592 700 Oe yc .
Oe 475 .
592 O yc O 700 .
Defuzzification Determine the moment value:
Determine the area:
358 ycOe250 358 ycOe250 ycA1 = O250 475Oe250 ycyccyc = 8346.
ya1 = O250 475Oe250 yccyc = 25.
ycA2 = O358 .
ycyccyc = 53352 ya2 = O358 .
yccyc = 112.
700 700Oeyc ycyccyc = 16277.
700Oe475 ycA3 = O592 700 700Oeyc yccyc = 25.
700Oe475 ya3 = O592 Thus, the center point of the fuzzy region is obtained as follows:
O yuN.
yc yccyc ycA ycA ycA yc = yc yuN.
yccyc = ya1 ya 2 ya 3 = Oyc 24 53352 16277.
= 475
Through manual calculations using the Mamdani method, the predicted cost of electricity consumption that must be paid is Rp475,000.
Figure 7.
The result of defuzzification mamdani method Copyright A 2025.
Author.
E-ISSN: 2621-086X https://doi.
org/10.
13057/ijas.
IJAS: Indonesian Journal of Applied Statistics.
Vol.
No.
Luthfia Zahra.
Mashuri Meanwhile, using the help of fuzzy Matlab software calculation of the Mamadani method .
resented in Figure .
, the predicted cost of electricity consumption that must be paid is Rp475,000.
Application of Sugeno Fuzzy Method .
Fuzzification While the Sugeno method is generally similar to the Mamdani method, the difference lies in the In Sugeno method, the output is a constant without complex defuzzification process.
Based on Figure 8, electricity consumption costs are categorized into LOW .
MEDIUM .
, and HIGH .
which are obtained based on the average output value of each category.
Figure 8.
Membership function of electricity consumption cost sugeno method .
Application of Implication Function, the Sugeno method uses the same basic fuzzy rules as the Mamdani method.
The rules used are presented in Table 3.
Rule Composition [R.
If the house area is MEDIUM and the number of electronic devices is NORMAL and the number of family members is FEW and the income is HIGH, then the electricity consumption cost is MEDIUM.
yuNycI.
= yuNycoycaycycAyayayaycOycA .
O yuNycycyyceycAycCycIycAyaya .
O yuNycycaycoyayaycO .
O yuNycyyayayaya .
= min .
uNycoycaycycAyayayaycOycA .
, yuNycycyyceycAycCycIycAyaya .
, yuNycycaycoyayaycO .
, yuNycyyayayaya .
) = min.
= 0.
Value yc39 = 470 .
btained from the average output value in the MEDIUM categor.
Defuzzification Defuzzification, the Sugeno method used is the weighted average method, which can be obtained as follows:
yc= .
u Oe ycyycyceycc39 O yc39 ) 0.
48 y 470
= 470
yu Oe ycyycyceycc39 Through manual calculations using the Sugeno method, the predicted cost of electricity consumption that must be paid is Rp470,000.
Copyright A 2025.
Author.
E-ISSN: 2621-086X https://doi.
org/10.
13057/ijas.
IJAS: Indonesian Journal of Applied Statistics.
Vol.
No.
Luthfia Zahra.
Mashuri Figure 9.
The result of defuzzification mamdani method Meanwhile, using the help of fuzzy Matlab software calculation of the Sugeno method .
resented in Figure .
, the predicted cost of electricity consumption that must be paid is Rp470,000.
Comparison of MAPE Values MAPE calculation using the Mamdani method.
ycUycn Oe yaycn Oc50 ycn=1 | ycU ycn ycAyaycEya = y 100% y 100% = 12.
306% OO 12.
MAPE calculation using the Sugeno method.
ycUycn Oe yaycn Oc50 ycn=1 | ycU ycn ycAyaycEya = y 100% y 100% = 9.
895% OO 9.
Based on the results of the calculations that have been carried out, the MAPE value of each method is obtained.
In the Mamdani method, the MAPE value is 12.
3%, which means that the accuracy level is This shows that the Mamdani method is well used to predict electricity consumption costs.
Meanwhile, the Sugeno method obtained a MAPE value of 9.
9% which means a accuracy level of 90.
is obtained.
Thus, the Sugeno method is proven to provide more accurate prediction results than the Mamdani method.
The accuracy results are presented in the following Table 4.
Table 4.
MAPE value accuracy result Calculation Method Mamdani Method Sugeno Method Copyright A 2025.
Author.
E-ISSN: 2621-086X MAPE Result Interpretation Good forecasting Highly accurate forecasting https://doi.
org/10.
13057/ijas.
IJAS: Indonesian Journal of Applied Statistics.
Vol.
No.
Luthfia Zahra.
Mashuri Comparison of Results Comparison of the results of the Mamdani method with the Sugeno method is presented in Table Table 5.
Comparison of MAPE results Respondent To- Actual Data (Rupia.
147,000
465,000
150,000
350,000
455,000
600,000
185,000
465,000
600,000
460,000
850,000
135,000
40,000
465,000
470,000
Mamdani Method Calculation (Rupia.
155,000 438,000 155,000 384,000 475,000 582,000 142,000 475,000 582,000 475,000
826,000
155,000
151,000
475,000
475,000
Sugeno Method Calculation (Rupia.
148,000 470,000 148,000 341,000 470,000 597,000 148,000 470,000 597,000 470,000
850,000
148,000
148,000
470,000
470,000
CONCLUSION
Based on the results of research on the application of fuzzy logic Mamdani and Sugeno methods in predicting electricity consumption costs in Margorejo Village.
Pati Regency, it is concluded that both methods can be used effectively with inputs in the form of house building area, number of electronic devices, number of family members, and income.
The Mamdani method involves the fuzzification process, implication function application using the min operator, rule composition with the max method, and defuzzification using the centroid method, resulting in 81 fuzzy rules.
Meanwhile, the Sugeno method uses a zero-order model with the output being a constant, defuzzification is done through weighted average.
The accuracy comparison results indicate that the Sugeno method achieves a lower MAPE value of 9.
9% compared to the Mamdani method, which records a MAPE of 12.
suggesting that the Sugeno method provides better predictive performance within the scope of this Therefore, the Sugeno method provides more accurate predictions.
However, this study has several limitations: .
the dataset is limited to a single village, which may affect the generalizability of the results.
the number of input variables is restricted and does not account for factors such as electricity tariff classes or usage behavior.
the Sugeno model used is limited to a zero-order The suggestions for further research are to add other relevant input variables, try other fuzzy methods as a comparison, and develop a prediction system in the form of mobile or web applications for ease of use.
Copyright A 2025.
Author.
E-ISSN: 2621-086X https://doi.
org/10.
13057/ijas.
IJAS: Indonesian Journal of Applied Statistics.
Vol.
No.
Luthfia Zahra.
Mashuri REFERENCES