Jurnal ICT : Information and Communication Technologies, 16 .
425-433 Published by: Marq & Cha Institute Jurnal ICT : Information and Communication Technologies Journal homepage: w.
id/index.
php/JICT Data Mining Using Multiple Linear Regression Method for Stock Prediction Methewkasly Pratama Sembiring 1.
Wiwi Verina2 Sistem Informasi.
Universitas Potensi Utama.
Medan.
Indonesia Article Info Article history Received : Oct 25, 2025 Revised : Oct 28, 2025 Accepted : Oct 30, 2025 Keywords:
Data Mining.
Inventory Prediction.
Multiple Linear Regression.
Website.
Abstract This study aims to apply data mining techniques using multiple linear regression to predict inventory at PT.
Sumber Jaya Motor's website.
today's digital era, companies face challenges in managing inventory, which can impact operational efficiency and customer satisfaction.
Therefore, accurate inventory prediction is essential to improve inventory management efficiency.
The multiple linear regression method was chosen because of its ability to link multiple independent variables with the dependent variable, thus providing more accurate predictions regarding required inventory.
The data used in this study includes information related to sales, suppliers, and demand obtained from PT.
Sumber Jaya Motor.
The results of the multiple linear regression application indicate that the developed model can provide inventory predictions with a high degree of accuracy.
This system is implemented on a website to facilitate real-time data-driven monitoring and decisionmaking.
With the implementation of this method, it is hoped that PT.
Sumber Jaya Motor can manage inventory more efficiently, reduce inventory costs, and improve customer service.
Corresponding Author:
Methewkasly Pratama Sembiring Sistem Informasi.
Universitas Potensi Utama.
Jl.
L Yos Sudarso KM 6.
5 Tj.
Mulia.
Medan, 20241.
Indonesia Email : asbiharis12@gmail.
This is an open access article under the CC BY-NC license.
Introduction Determining stock levels is crucial in any business process, specifically meeting the high demand for vehicle spare parts.
The demand for spare parts is currently quite high due to the annual increase in vehicle usage.
One example is PT.
Sumber Jaya Motor, located at Jl.
William Iskandar/Pancing No.
PT.
Sumber Jaya Motor generates relatively high daily sales volumes, significantly impacting warehouse inventory.
The impact of product availability is crucial, particularly for the most soughtafter and best-selling spare parts.
PT.
Sumber Jaya Motor is a company engaged in the sales and distribution of automotive spare In carrying out its business operations, the company relies on stable and accurate inventory availability to optimally meet customer demand.
However, in practice, inventory management often encounters challenges such as overstock or stockouts, which directly impact operational efficiency and customer satisfaction.
One of the main causes of these problems is a suboptimal inventory forecasting and management Homepage: w.
id/index.
php/JICT JICT p-ISSN 2086-7867 e-ISSN 2808-9170 Until now, the process of planning the quantity of goods to be stored in warehouses has been done manually and based on rough estimates based on previous experience, without using a systematic, data-driven approach.
This makes it difficult for companies to anticipate market demand and sales trends for specific products.
With the advancement of information technology, particularly in the field of data mining, companies can utilize data analysis techniques to uncover patterns and relationships in historical sales One method that can be used for prediction is multiple linear regression.
This method allows companies to analyze the influence of several variables on inventory, such as previous sales volume, seasonal trends, prices, and product types, thus allowing for more accurate stock requirement The implementation of the multiple linear regression method in a website-based information system is an effective solution because it allows for automated data management and prediction processes, in real time, and is easily accessible by warehouse management and marketing departments.
The website-based system also provides convenience in monitoring and decision-making, without having to rely on a specific physical location.
With this data mining-based stock prediction system.
PT.
Sumber Jaya Motor is expected to improve inventory management efficiency, reduce the risk of losses due to estimation errors, and increase customer satisfaction through better product availability.
Therefore, this research is important to design and implement a website-based stock prediction system using an integrated and effective multiple linear regression method.
Research Methodolgy In developing the system, the author uses the waterfall model or software life cycle, the software life cycle has the following stages:
Data collection System Creation Program Code Writing Program Testing Maintenance Result Figure 1.
Waterfall Diagram The Waterfall method has several stages in its development: requirements analysis, system design, coding, program testing, and system maintenance.
Data Collection The researcher collected data containing the elements that must be included in the design to solve the existing problems and meet the objectives.
The data required for system design included Data Mining Using Multiple Linear Regression Method for Stock Prediction (Methewkasly Pratama Sembiring, et a.
A p-ISSN 2086-7867 e-ISSN 2808-9170 product stock prediction data, user data, and the programming language used to create the application.
PHP.
The author conducted a field study by conducting direct field visits to collect data, including direct visits to the study location.
The data collection techniques used by the author Observation The author observed product data, stock data, and sales data on inventory.
Interview This technique involved direct face-to-face meetings with relevant parties to obtain clarification on previously unclear issues, including the system mechanisms used in the company, and to ensure that the data collected was accurate.
The interview process was conducted with Mrs.
Imelda Aritonang, a sales representative, as follows:
What are the challenges frequently encountered in inventory management at PT.
Sumber Jaya Motor? Answer: We often experience overstock or understock because demand forecasts are still done manually, based solely on experience or previous sales data without in-depth analysis.
Does the company currently use a system to monitor or predict inventory? Answer: We currently use an Excel-based inventory recording system.
However, there is no system capable of predicting inventory needs automatically or based on historical data.
What is the impact of forecast errors on company operations? Answer: The impact is significant.
If inventory runs out, customers cannot be served properly.
Conversely, if inventory is overstocked, inventory piles up in the warehouse, increasing storage .
Is previous sales and inventory data available and well-documented? Answer: Yes, we have documented monthly sales data and incoming/outgoing inventory data for the past several years, but it has not been fully utilized for analysis.
What would you think if a stock forecasting system could be automated using website-based Answer: It would be very helpful, as such a system would speed up the decision-making process, reduce errors, and be accessible from anywhere.
We hope to have a system that is user-friendly and accurate.
What variables or factors do you think influence inventory requirements? Answer: Some important factors include: previous sales trends, seasonality .
ecause some spare parts are in demand at certain time.
, selling price, type of item, and sometimes market conditions such as promotions or discounts from manufacturers.
Does the company have an internal IT team for system development? Answer: We don't have an internal IT team.
We typically collaborate with third parties or external developers if we want to create a technology-based system.
So far, what is the company's procurement and inventory recording workflow? Answer: The workflow begins with recording daily sales, then at the end of each month, warehouse staff conducts a stock recap.
After that, the purchasing department orders goods from distributors based on the estimated stock requirements for the following month.
What are your expectations for the inventory forecasting system that will be developed? Answer: We hope this system can help estimate stock accurately, reduce the risk of stockouts or stockpiling, and support the company's overall operational efficiency.
System Development In general, the Predictive Information System uses the Multiple Linear Regression Method and uses Unified Modeling Language design models, namely use case diagrams, class diagrams, activity diagrams, and sequence diagrams.
Method The author chose the Multiple Linear Regression method to design sales estimates using the Linear Regression method.
This method is one of the most important approaches in engineering for: .
JICT.
Vol.
No.
October 2025: 425-433
JICT
p-ISSN 2086-7867 e-ISSN 2808-9170 regression or equation formation from discrete data points .
n modelin.
, and .
measurement error analysis .
n model validatio.
Program Testing At this stage, comprehensive application testing is conducted, including functional testing and system robustness testing.
Black-box .
testing is software testing that tests the application's functionality against its internal structure or operation.
Maintenance Software that is difficult to deliver to system users will inevitably undergo changes.
These changes can occur due to errors as the software must adapt to the new environment.
Results The researcher used the system to predict inventory using data mining techniques using the Multiple Linear Regression method.
Results and Discussion The author will forecast the sales of Shock Absorber vehicle spare parts for the period January 2020 to December 2024.
Sales data can be seen in Table 1.
Product name Bantalan Shok Bantalan Shok Bantalan Shok Bantalan Shok Bantalan Shok Bantalan Shok Bantalan Shok Bantalan Shok Bantalan Shok Bantalan Shok Bantalan Shok Bantalan Shok Bantalan Shok Bantalan Shok Bantalan Shok Bantalan Shok Bantalan Shok Bantalan Shok Bantalan Shok Bantalan Shok Bantalan Shok Bantalan Shok Bantalan Shok Bantalan Shok Bantalan Shok Bantalan Shok Bantalan Shok Bantalan Shok Bantalan Shok Bantalan Shok Bantalan Shok Bantalan Shok Bantalan Shok Bantalan Shok Bantalan Shok Bantalan Shok Table 1.
Shock Absorber Sales Data .
Month Year Sold Initial Stock Januari Februari Maret April Mei Juni Juli Agustus September Oktober November Desember Januari Februari Maret April Mei Juni Juli Agustus September Oktober November Desember Januari Februari Maret April Mei Juni Juli Agustus September Oktober November Desember Product in Data Mining Using Multiple Linear Regression Method for Stock Prediction (Methewkasly Pratama Sembiring, et a.
A p-ISSN 2086-7867 Product name Bantalan Shok Bantalan Shok Bantalan Shok Bantalan Shok Bantalan Shok Bantalan Shok Bantalan Shok Bantalan Shok Bantalan Shok Bantalan Shok Bantalan Shok Bantalan Shok Bantalan Shok Bantalan Shok Bantalan Shok Bantalan Shok Bantalan Shok Bantalan Shok Bantalan Shok Bantalan Shok Bantalan Shok Bantalan Shok Bantalan Shok Bantalan Shok Month Januari Februari Maret April Mei Juni Juli Agustus September Oktober November Desember Januari Februari Maret April Mei Juni Juli Agustus September Oktober November Desember Year Sold Initial Stock e-ISSN 2808-9170 Product in Based on the explanation of the table above, the results of the Constant Value and Regression Coefficient are as follows:
Month Year Januari Februari Maret April Mei Juni Juli Agustus September Oktober November Desember Januari Februari Maret April Mei Juni Juli Agustus September Table 2.
Constant Values and Regression Coefficients Stock Product Sold (X.
X1*Y X2*Y X1*X2 (X.
in (Y) JICT.
Vol.
No.
October 2025: 425-433
X1^2
X2^2
Y^2
JICT
Month p-ISSN 2086-7867 Year Oktober November Desember Januari Februari Maret April Mei Juni Juli Agustus September Oktober November Desember Januari Februari Maret April Mei Juni Juli Agustus September Oktober November Desember Januari Februari Maret April Mei Juni Juli Agustus September Oktober November Desember Rata-rata Total n = 60 Sold (X.
Stock (X.
X1^2
X2^2
Y^2
e-ISSN 2808-9170
Product
X1*Y
X2*Y
X1*X2
in (Y) Based on Table 2, the following equation is obtained using the multiple linear regression method:
Equation I OcYA = OcYA - n YA OcYA = 929085 - .
* .
* .
) OcYA = 929085 Ae 60 * 13456 Data Mining Using Multiple Linear Regression Method for Stock Prediction (Methewkasly Pratama Sembiring, et a.
A p-ISSN 2086-7867 e-ISSN 2808-9170 OcYA = 929085 Ae 807360 OcYA = 121725 Equation II OcX1A = OcX1A - n X1A OcX1A = 1292326 - .
* .
* .
) OcX1A = 1292326 - 60 * 18769 OcX1A = 1292326 - 1126140 OcX1A = 166186 Equation i OcX2A = OcX2A - n X2A OcX2A = 1369255 - .
* .
* .
) OcX2A = 1369255 - 60 *19321 OcX2A = 1369255 - 1159260 OcX2A = 209995 Equation IV OcX1Y = OcX1Y - n X1Y OcX1Y = 970045 - .
* .
* .
) OcX1Y = 970045 - 60 * 15892 OcX1Y = 970045 - 953520 OcX1Y = 16525 Equation V OcX2Y = OcX2Y - n X2Y OcX2Y = 940148 - .
* .
* .
) OcX2Y = 940148 - 60 * 16124 OcX2Y = 940148 - 967440 OcX2Y = -27292 Equation VI OcX1X2 = OcX1X2 - n X1X2 OcX1X2 = 1148258 - .
* .
* .
) OcX1X2 = 1148258 - 60 * 19043 OcX1X2 = 1148258 - 1142580 OcX1X2 = 5678 Thus, the results obtained for the constant value a and the regression coefficients b1 and b2 are as b1 = (OcX2A)(OcX1Y)-(OcX1X.
(OcX2Y) / (OcX1A)(OcX2A)-(OcX1X.
2 b1 = .
995*16.
8*-27.
/ .
186*209.
2 b1 = .
0167375 Ae (-154963.
/ 34898229070 - 32239684 b1 = 3625131351 / 34865989386 b1 = 0.
b2 = (OcX1A)(OcX2Y)-(OcX1X.
(OcX1Y) / (OcX1A)(OcX2A)-(OcX1X.
2 b2 = .
186*-27.
8*16.
/ .
186*209.
2 b2 = -4535548312 Ae 93828950 / 34898229070 Ae 32239684 b2 = -4629377262 / 34865989386 b2 = -0.
a = Y - b1X1 - b2X2 a = 116 - .
10397328212506 * .
- (-0.
13277630560683 * .
a = 116 Ae 14,24433936 Ae (-18,4559.
a = 120.
21156682822 = 120
So that the regression equation model is obtained from the results of the calculations in the case above, it can be seen that the results of the stock forecast for Shock Bearings for the 2025 period are as follows:
JICT.
Vol.
No.
October 2025: 425-433
JICT
p-ISSN 2086-7867 e-ISSN 2808-9170 Y = a b1X1 b2X2 .
Month Januari Februari Maret April Mei Juni Juli Agustus September Oktober November Desember Year Y = a b1X1 b2X2 Y = 120 .
(-0.
Y = 120 .
(-0.
Y = 120 .
(-0.
Y = 120 .
(-0.
Y = 120 .
(-0.
Y = 120 .
(-0.
Y = 120 .
(-0.
Y = 120 .
(-0.
Y = 120 .
(-0.
Y = 120 .
(-0.
Y = 120 .
(-0.
Y = 120 .
(-0.
Forecast Amount 91720886535 atau dibulatkan = 122 54363718219 atau dibulatkan = 117 659702656 atau dibulatkan = 121 62310282469 atau dibulatkan = 119 12025202167 atau dibulatkan = 122 1931017307 atau dibulatkan = 107 14296923531 atau dibulatkan = 119 30687335224 atau dibulatkan = 125 52692635039 atau dibulatkan = 121 62310282469 atau dibulatkan = 119 41812756078 atau dibulatkan = 104 00576446364 atau dibulatkan = 122 Conclusion Based on the implementation of data mining using the Multiple Linear Regression method for predicting spare part inventory at PT.
Sumber Jaya Motor, it can be concluded that the developed webbased system significantly enhances the accuracy and efficiency of stock management.
By utilizing historical sales data and relevant variables, the system successfully reduces errors in determining the required stock levels, enabling the company to anticipate future demand more effectively.
The system, built using PHP and a MySQL database, also simplifies the process of recording, monitoring, and updating inventory, allowing users to work more quickly and with fewer mistakes.
This improvement not only increases operational efficiency but also helps the company address recurring challenges in forecasting spare part needs for upcoming periods.
Furthermore, this research highlights several opportunities for future development.
The system may be enhanced by integrating additional forecasting models such as ARIMA.
Random Forest, or Artificial Neural Networks to compare and improve prediction accuracy.
Incorporating external variables such as seasonal trends, market fluctuations, and supplier data could further strengthen the forecasting results.
The addition of interactive analytical dashboards, automatic stock-level notifications, and a mobile-based application would also expand usability and support real-time decision-making.
Overall, the designed system provides a strong foundation for continuous improvement in inventory forecasting and management.
References