p–ISSN: 2723 - 6609 e-ISSN: 2745-5254
Vol. 5, No. 7 July 2024 http://jist.publikasiindonesia.id/
Jurnal Indonesia Sosial Teknologi, Vol. 5, No. 7, July 2024 3392
Shallot Production Prediction System Using the C.45 Decision
Tree Algorithm
Aghnie Kurnia Fadhila
Universitas Jendral Achmad Yani, Indonesia
Email:
*Correspondence
ABSTRACT
Keywords: C4.5
algorithm; shallot;
decision tree; data mining.
This research applies the C4.5 algorithm, which is a machine
learning algorithm for classification using decision trees, in
a case study for predicting the performance of shallot
production. The data used includes attributes such as
production yield, land area, and productivity. The C4.5
Decision Tree algorithm is utilized to build an accurate
prediction model after going through data cleaning and
training processes. This study results in an application that
can perform the entire process of initial data processing to
data analysis using the aforementioned technique, making it
efficient and effective in analyzing large amounts of data to
obtain optimal prediction results.
Introduction
Shallots (Allium ascalonicum L) are vegetables that have high commercial value
both in terms of economy and nutrition (Damayanti, 2022). The health benefits of shallots
have been widely felt, and the related industry is booming, leading to an increase in
demand in the domestic market. (Nurhayati, Sibuea, Kusbiantoro, Silaban, & Wanto,
2022) The demand for shallots in Indonesia, both as a vegetable and as a seed, continues
to increase by 5% every year as the population grows and consumer interest increases
(Baihaqi, Handayani, & Pujianto, 2019).
Shallot production is one of the crucial agricultural sectors for the Indonesian
economy. Shallots, as one of the main horticultural commodities, have a significant role
in meeting the food needs of the community and making an important contribution to
farmers' income (Priyaungga, Aji, Syahroni, Aji, & Saifudin, 2020). However,
fluctuations in production caused by various factors such as climate change, pest attacks,
and suboptimal cultivation techniques are often a major challenge for farmers and
stakeholders (Hana, 2020).
To overcome these problems, a production prediction system is needed that can
provide accurate and reliable information. With a prediction system, farmers can better
plan their cultivation activities, optimize the use of resources, and minimize the risk of
losses due to production that is not by estimates (Kesuma & Kholifah, 2019). Decision
Tree C4.5 works by dividing the dataset into several subsets based on the attributes that
are most significant in influencing the target variable, in this case, the production of
Shallot Production Prediction System Using the C.45 Decision Tree Algorithm
Jurnal Indonesia Sosial Teknologi, Vol. 5, No. 7, July 2024 3393
shallots. Each branch in the decision tree represents a specific condition or decision that
leads to the outcome of the prediction.
Based on the trend of increasing demand for shallots in the domestic market, it is
important to analyze this problem to predict production that is likely to increase or
decrease (Wajhillah & Yulianti, 2017). Previous research (Zulkarnain & Marciano, 2022)
has been conducted on the Prediction of Shallot Harvest Production Using a Simple
Linear Regression Method. The results show a Mean Squared Error (MSE) of 2073311,
a Root Mean Squared Error (RMSE) score of 45533, and an R2 score of 0.98% or 98%
accuracy (Ghozali & Wibowo, 2019). Although previous studies predicted increased
production, evaluations with different analysis methods could provide a new perspective
on accuracy. Therefore, this study will compare the prediction results with previous
studies to determine significant differences in accuracy or other factors that need to be
considered (Maulana, Martanto, & Ali, 2023).
Research Methods
This research involves several stages in the data mining process. First, data was
included as a research subject. Next, preprocessing is carried out to prepare the data.
Then, the prediction process is carried out by applying the C4.5 Decision Tree algorithm
for testing. Finally, the results are analyzed and reported in the form of publications. The
diagram in Figure 1 shows the overall stages of the research.
Figure 1 Description of the research flow
Data Collection
In this study, data was obtained from BPS (Central Statistics Agency).
Aghnie Kurnia Fadhila
Jurnal Indonesia Sosial Teknologi, Vol. 5, No. 7, July 2024 3394
Pre-Processing
At this stage, the existing data is transformed into a suitable format to be used as an
object in the research. There are two stages of pre-processing carried out in this study,
namely Data Cleaning and Data Selection.
1. Data Cleaning
Data Cleaning is an important first step to maintaining the quality of data from non-
compliant data sources. This step involves correcting and removing inappropriate data to
make the information obtained more relevant. Incomplete, redundant, or inconsistent data
will be eliminated to make the analysis process easier.
2. Data Selection
The second stage after Data Cleaning is Data Selection, where data that already has
complete information is selected according to the needs of the necessary information. In
this stage, the data is selected from the overall 8 required attributes.
Prediction Process using Decision Tree C4.5
At this stage, data derived from the latest BPS and assessment sources that have
been processed into a data set during the previous pre-processing stage will be examined.
The process and steps of this method are as follows:
Data Training
Record data from BPS and assessments that have gone through the pre-processing
stage and combined into one dataset will be used as training data to train the C4.5 decision
tree algorithm. It aims to produce an accurate model to make predictions.
Algoritma Decision Tree C4.5
In this stage, the rules of the C4.5 Decision Tree algorithm are used to form a
decision tree model based on training data.
Decision Tree Model Creation
The decision tree model is made based on the classification results using the C4.5
Decision Tree algorithm, by the predetermined tree formation rules.
Data Testing
The data is used to test the performance of a previously trained algorithm when
faced with new data that has never been seen before. The test data will be classified using
the tree model that has been created.
Decision Tree Model Performance Testing
Testing is carried out on the decision tree model by entering test data into the tree
model that has been created, to evaluate the performance of the algorithm that has been
trained.
Labelling Data Uji
Once the test data has been classified using the prediction model, the next step is to
label the test data results between the categories of "less", "very", and "adequate".
Software Design and Development
This stage of software design and development is carried out by applying a
prediction tree model to the software creation process.
Testing and Evaluation
Shallot Production Prediction System Using the C.45 Decision Tree Algorithm
Jurnal Indonesia Sosial Teknologi, Vol. 5, No. 7, July 2024 3395
The evaluation is carried out by calculating the level of accuracy and assessing the
extent to which the system has succeeded in generating correct information based on the
model that has been created.
Results and Discussion
Analysis of Prediction of Shallot Production Using Decision Tree C4.5 Algorithm
Analysis using the Decision Tree C4.5 algorithm was carried out to predict the yield
of shallots based on historical data and factors that affect it such as weather, soil type, and
cultivation techniques. The methodology includes data collection, data cleaning, sharing
of training and test data, C4.5 model development, model evaluation, optimization, and
interpretation of results. This analysis aims to identify the main factors that affect shallot
production so that it can help farmers and stakeholders in making decisions related to
agricultural practices to increase production efficiently.
Data Collection
In this study, secondary data is taken from various websites to find data that is the
focus of the research. The purpose of these efforts is to ensure that the data used is valid
and supports the smooth and successful conduct of the research. After various search
efforts, finally, the dataset used consisted of 243 data records taken from the Central
Statistics Agency (BPS). By using this dataset, it is hoped that research can be carried out
well and produce accurate and relevant results.
Shallot Dataset
The following is a view of the dataset taken from the Central Statistics Agency with
the number of 8 attributes which can be seen in Table 1 of the Shallot Daset.
Aghnie Kurnia Fadhila
Jurnal Indonesia Sosial Teknologi, Vol. 5, No. 7, July 2024 3396
Data Cleaning
In the data cleaning stage, identification is an important step to ensure the quality
of the data used in the analysis. Therefore, in this study, the data-cleaning process is
carried out by identifying empty values. In the Shallot data obtained at the Central
Statistics Agency, the dataset has been checked that there is a missing value in the data.
As well as the removal of attributes on the shallot dataset. By carrying out this data-
cleaning process, the data used in the analysis becomes more accurate and reliable. In
Id
Provincial
Code
Name
Provinsi
District
Code
Kota
Broad
Onion
Production
Red
Year
Parameter
1
32
West
Java
3
20
10
2013
Less
2
32
West
Java
2
24
285
2013
Very
3
32
West
Java
2
31
183
2013
Enough
4
32
West
Java
1
2915
31682
2013
Very
5
32
West
Java
3
1967
19728
2013
Very
6
32
West
Java
1
13
90
2013
Enough
7
32
West
Java
2
0
0
2013
Less
8
32
West
Java
3
237
2218
2013
Enough
9
32
West
Java
3
3658
36449
2013
Enough
10
32
West
Java
2
2150
23683
2013
Very
11
32
West
Java
1
27
204
2013
Enough
12
32
West
Java
1
197
950
2013
Less
13
32
West
Java
2
0
0
2013
Less
14
32
West
Java
3
0
0
2013
Less
15
32
West
Java
1
0
0
2013
Less
16
32
West
Java
2
3
21
2013
Enough
17
32
West
Java
2
15
83
2013
Enough
18
32
West
Java
3
0
0
2013
Less
19
32
West
Java
3
0
0
2013
Less
20
32
West
Java
1
0
0
2013
Less
Shallot Production Prediction System Using the C.45 Decision Tree Algorithm
Jurnal Indonesia Sosial Teknologi, Vol. 5, No. 7, July 2024 3397
addition, the proper data cleaning process also ensures that the results of the analysis
produced are of good quality and can be accounted for.
City district
code
Broad
Shallot production
Parameter
3
20
10
less
2
24
285
very
2
31
183
enough
1
2915
31682
very
3
1967
19728
very
1
13
90
enough
3
237
2218
enough
3
3658
36449
enough
2
2150
23683
very
1
27
204
enough
1
197
950
less
2
3
21
enough
2
15
83
enough
Transformation Data
In this data analysis process, Data Transformation is carried out to facilitate data
processing and analysis more effectively. Data Transformation is carried out by changing
data variables into numerical data forms.
The dataset used in this study consists of 8 attributes which include ID, province
code, province name, city district code, Area, shallot production, year, and Parameters.
Of the datasets, 6 attributes have numerical data types, while the other two attributes have
nominal data types. To overcome this, this study transforms data by converting two
nominal attributes into numerical ones.
The nama_provinsi attribute and the Parameter attribute which originally consisted
of less, sufficient, and very were also changed to numeric by replacing less with a value
of 0, sufficient with a value of 1, and very with a value of 2. This altered data is then used
in research to facilitate analysis and testing.
Table 3
Transformation Data
City district code
Broad
Shallot production
Parameter
3
20
10
0
2
24
285
2
2
31
183
1
1
2915
31682
2
3
1967
19728
2
1
13
90
1
3
237
2218
1
Aghnie Kurnia Fadhila
Jurnal Indonesia Sosial Teknologi, Vol. 5, No. 7, July 2024 3398
3
3658
36449
1
2
2150
23683
2
1
27
204
1
1
197
950
0
2
3
21
1
2
15
83
1
In Table 3. Data Transformation is an example of a dataset that has been pre-
processed with Data Cleaning and Data Transformation, the dataset is ready to be carried
out for the Prediction process with the C4.5 Model.
Prediction Process with Model C4.5
Of all the data generated through preprocessing, this study divided the data into two
parts, namely training and testing data, where 171 or 70% was partitioned for training
data and 72 or 30% for testing data.
Table 4
Proportion of each class
Status
Code
Kabupaten_kota
Broad
Shallot
production
Parameter
Sum
Propos
ition
Sum
Prop
ositio
n
Sum
Propos
ition
Sum
Prop
ositio
n
Less
90
0.37
73
0.30
56
0.23
61
0.25
Enough
66
0.27
85
0.35
134
0.55
61
0.25
Very
88
0.36
85
0.35
53
0.22
121
0.50
Total(s)
243
1.00
243
1.00
243
1.00
243
1.00
(Parameter 0,1,2) =
1.557
1. Calculating Information Gain
This calculation is intended for each attribute used with S as a class set of less,
sufficient, and very much. Classes are less with code 0, classes are sufficient with ode 1
and classes are very with code 2 s reobtain. Cap
Table 5
Entropy (City District Code)
S
City district
code
Less
Enough
Very
S1
0
23
15
31
S2
1
37
32
29
Shallot Production Prediction System Using the C.45 Decision Tree Algorithm
Jurnal Indonesia Sosial Teknologi, Vol. 5, No. 7, July 2024 3399
S3
2
30
18
28
Calculating Retro Gain
From the calculation process, the result of the information gain of the
kode_kabupaten_kota attribute was -1,471. The calculation of information acquisition is
calculated for all attributes, so the results are shown in Table 6.
Table 6
Attribute calculation results
No.
Attribute
Info Gain
Split Info
Gain Ratio
1.
Code-
kabupaten_kota
-2.319
1.557
-1.471
2.
Broad
-0.012
0.981
-0.012
3.
Produksi_bawan
g_merah
- 1.739
1.482
-1.735
Based on the construction of the model that has been made, a tree model is created
that depicts the relationship between attributes.
Aghnie Kurnia Fadhila
Jurnal Indonesia Sosial Teknologi, Vol. 5, No. 7, July 2024 3400
Figure 2
Tree model
The results of the decision tree algorithm for the classification of shallot production
show that shallot production can be predicted using the parameters of land area and
production amount. This decision tree classifies production into three categories: less,
adequate, and very. In this tree, the main variable that affects the classification is the
amount of shallot production, with the initial node dividing the data based on whether the
shallot production is less than or equal to 1.5. This node has a high entropy, indicating
great uncertainty in the early stages. The division continues with additional parameters
such as land area, which helps to clarify the classification. For example, shallot
production below 1.5 is generally classified as "poor", while production above 1.5 but
below 4013.5 is generally classified as "adequate". Overall, this model shows that the
increase in the amount of shallot production and the variation in land area significantly
affect the classification of shallot production levels.
Conclusion
The above problem about the data of shallot prediction results can be solved by the
C4.5 algorithm method using classification rules to develop a prediction model that can
predict the quality of shallots. And identify important attributes that affect the quality of
shallots.
Shallot Production Prediction System Using the C.45 Decision Tree Algorithm
Jurnal Indonesia Sosial Teknologi, Vol. 5, No. 7, July 2024 3401
Bibliography
Baihaqi, Dimas Imam, Handayani, Anik Nur, & Pujianto, Utomo. (2019). Perbandingan
metode Naive Bayes dan C4. 5 untuk memprediksi mortalitas pada peternakan ayam
broiler. Simetris: Jurnal Teknik Mesin, Elektro Dan Ilmu Komputer, 10(1), 383–390.
Damayanti, Desi. (2022). Implementasi Algoritma C4. 5 Prediksi Produksi Komoditas
Tanaman Perkebunan Berdasarkan Luas Lahan. Tin: Terapan Informatika
Nusantara, 2(10), 571–579.
Ghozali, Muhammad Rizal, & Wibowo, Rudi. (2019). Analisis Risiko Produksi
Usahatani Bawang Merah di Desa Petak Kecamatan Bagor Kabupaten Nganjuk.
Jurnal Ekonomi Pertanian Dan Agribisnis, 3(2), 294–310.
Hana, Fida Maisa. (2020). Klasifikasi Penderita Penyakit Diabetes Menggunakan
Algoritma Decision Tree C4. 5. Jurnal SISKOM-KB (Sistem Komputer Dan
Kecerdasan Buatan), 4(1), 32–39.
Kesuma, Chandra, & Kholifah, Desiana Nur. (2019). Sistem Informasi Akademik
Berbasis Web Pada Lkp Rejeki Cilacap. EVOLUSI: Jurnal Sains Dan Manajemen,
7(1).
Maulana, Alfin, Martanto, Martanto, & Ali, Irfan. (2023). Prediksi Hasil Produksi Panen
Bawang Merah Menggunakan Metode Regresi Linier Sederhana. JATI (Jurnal
Mahasiswa Teknik Informatika), 7(4), 2884–2888.
Nurhayati, Nurhayati, Sibuea, Mhd Buhari, Kusbiantoro, Dedi, Silaban, Martina, &
Wanto, Anjar. (2022). Implementasi Algoritma Resilient untuk Prediksi Potensi
Produksi Bawang Merah di Indonesia. Building of Informatics, Technology and
Science (BITS), 4(2), 1051–1060. https://doi.org/10.47065/bits.v4i2.2269
Priyaungga, Bayu Aji, Aji, Dwi Bayu, Syahroni, Mukron, Aji, Nurul Tri Sukma, &
Saifudin, Aries. (2020). Pengujian Black Boxpada Aplikasi Perpustakaan
Menggunakan Teknik Equivalence Partitions. Jurnal Teknologi Sistem Informasi
Dan Aplikasi ISSN, 2654, 3788.
Wajhillah, Rusda, & Yulianti, Ita. (2017). Penerapan algoritma c4. 5 untuk prediksi
penggunaan jenis kontrasepsi berbasis web. Klik-Kumpul. J. Ilmu Komput, 4(2), 160.
Zulkarnain, Muhammad, & Marsisno, Waris. (2022). Penerapan Pembelajaran Mesin
Untuk Estimasi Luas Lahan Bawang Merah Berdasarkan Data Citra Satelit Resolusi
Menengah. Seminar Nasional Official Statistics, 2022(1), 1005–1016.
https://doi.org/10.34123/semnasoffstat.v2022i1.1307
