Decision Support System for Determining the Feasibility of Material Utilization Using the Naïve Bayes Method

A company is engaged in the mechanical and electrical industry in providing goods and services, especially in medium and high power installations. In this company, the data management process carried out is still semi-computer, on a sheet of paper then copied into a computer with the Microsoft Excel application. Every work done by a technician in using material contains materials that are not used up or used materials. The material is then returned to the warehouse to be used as a report which is then carried out by the utilization process. The purpose of this research is to create a system that can assist in providing suggestions for used or used materials in the warehouse for utilization by applying the Naïve Bayes method. Naïve Bayes is a data mining technique with a classification function. This data mining application methodology uses the KDD (Knowledge Discovery in Database) stage starting from the selecting, preprocessing, transformation, data mining, and evaluation stages. The testing technique used for data validation uses the k-fold cross-validation technique accompanied by a Confusion Matrix. This research has produced a decision support system application and has been tested with an accuracy rate of 73.33%.


INTRODUCTION
The development of information technology which is getting faster nowadays makes companies use this technology to produce a system capable of processing data in various ways to obtain quality information so that the work process can be more relevant, accurate, and timely in helping to improve performance and strengthen competitiveness company.
PT. Nusa Abadi Elektrika is a company engaged in the mechanical and electrical industry providing goods and services, especially in medium and high power installations. The first problem is that in this company the data management process carried out is still semi-computer, on a sheet of paper which is then copied by the officer into a computer with the Microsoft Excel application.
Every use of the material in doing work done by technicians usually contains materials that are not used up or used materials. The material will then be returned to the warehouse for recording which will be used as a material usage report then the utilization process is carried out.
Along with the large number of repair work, this creates a second problem in the part of the utilization process which runs slowly and has the potential to be inconsistent and often miscommunication occurs, causing the accumulation of unused material items in the warehouse which makes the room narrow considering the limited space conditions.
With the existence of a decision support system to determine the feasibility of using the material using the Naive Bayes method, it will make it easier for officers at the warehouse to classify materials and provide suggestions on the feasibility of the material to be used.
The scope of this system is to act as decision support in the warehouse section in providing suggestions to determine the feasibility of used and unused materials in the warehouse of PT. Nusa Abadi Elektrika with decision-making analysis techniques in determining the feasibility of the material using the Naïve Bayes method. The purpose of this system is to provide suggestions for material utilization and maximize the use of materials in the work process. The benefits of this system are obtaining accurate and effective and efficient information in material management and minimizing errors in data collection that are still carried out conventionally.
In making this research, it is necessary to carry out a literature study to serve as a reference for existing studies, including: This research to implementing the naïve bayes method that measuring of feasibility utilization on decision support system.

METHOD
The model used in this study follows the stages in the KDD (Knowledge Discovery in Database) process. According to EforieBuulolo (2020: 1-5) [9], KDD is a process of discovery in a database or the entire process of extraction or identification of patterns, knowledge, and potential information from large data sets. Knowledge and information generated from KDD are valid, new, easy to understand, useful.
The stages of the Knowledge Discovery in Database process, see the figure 1 as bellow: Not all existing data can be used. Therefore, the data selected. The data selection activity includes creating a data set on file storage.
b. Pre-processing / Cleaning At this stage, the data that has been selected will be cleaned. The cleaning process includes removing duplicate data, correcting inconsistent data, and correcting data errors. In preprocessing/cleaning, the process of enriching data can also be carried out by adding other relevant information which is called enrichment.
c. Transformation In data mining, there are so many algorithms/ methods/techniques that can be used. For the KDD process, it is changed first according to the algorithms/methods/techniques used in data mining.
d. Data Mining Data Mining is the main stage in KDD. Data mining is the process of extracting and searching for useful knowledge and information using certain algorithms/methods/ techniques by the knowledge or information sought.

e. Interpretation / Evaluation
Knowledge or information generated from the data mining process will be presented or displayed in a form that is easily understood by interested parties such as information displayed in the form of graphics, decision trees, in the form of rules. The knowledge or information generated from the data mining process is examined whether or not it contradicts previous facts or hypotheses.

A. Selection
The source of the dataset used in this study is used material and used materials in the warehouse. The material comes from work residue under various conditions of damage. Meanwhile, the feasibility data comes from the warehouse head's decision that has been approved by the company leadership.

B. Pre-Processing Data
To find out the feasibility of each material, it is checked again manually in the warehouse after which it is evaluated with the leader for the next decision. As for some materials whose decisions have not been approved by the company leadership. Examples of data collection are arranged in table 1 Preprocessing Data.

C. Transformation
To make it easier for the next mining process, the data is transformed according to the type of data at the transformation stage, where the damage conditions will be grouped into categorical data. The following are the results of the transformed dataset in table 2 Transformation.

D. Data Mining
In data processing, the modeling stage is carried out for the classification process, namely applying the Naïve Bayes algorithm with Rapidminer software as shown in figure 2 below. This Naive Bayes calculation method is a method used in finding the probability where the calculation is carried out by the system. In calculating the naïve Bayes method, training data or system learning data are needed. The data amounted to 30 material data for calculations carried out by researchers. The following provides an example of the data to be tested (look at table 3 Data Testing).   Class can be looked at table 5, table 6, and table 7 as below:  e. Evaluation Evaluation is carried out using the K-Fold Cross-validation technique (see Fig 3 below) with the value of k taken 5 fold so that from 30 data will be 6 data subsets with the same size, which is about 5 data. From each of the 6 subsets, 25 data becomes training data and 5 data becomes test data. The training and testing process model can be seen in figure 2 above. The output of the performance results from this algorithm is a classification of materials that are included in the classification of the results "feasible" or "not" which is applied to the testing data. The amount of data that are predicted correctly by the Naïve Bayes algorithm is shown in  Predictions are not feasible 4 13 Table 8 Confusion Matrix Naïve Bayes algorithm is: i. The actual amount of DECENT and predicted DECENT data is 9 ii.
The actual amount of data that DECIDES and predicts NOT is 4 iii.
The actual number of data that DID NOT and predictably WORTHY is 4 iv.
The actual number of data that is NOT and predicted NOT is 13 So that the evaluation for this model uses the accuracy and error values.
The accuracy of the model is: The results of the implementation of the display interface for each menu of the decision support system for determining the feasibility of material used with the naïve Bayes method are as follows: a. The Login Menu Display The login menu display (see figure 4) appears if you have logged in to the web browser. The user must enter a user name and password first. After a successful login, it will be directed to the Home or Dashboard menu (see figure 5) as the initial display of the system. In this section of the menu (see figure 6) is a database or reference data that will be used as test material to classify testing data.    figure 8) is a report on the results of data testing that has been carried out on the data testing menu.

CONCLUSION
Based on the conclusions from the discussion topic of the Decision Support System for Determining the Feasibility of Material Utilization with the Naïve Bayes Method as follows: 1) Currently, the obstacles or problems faced by warehouse officers are in determining the appropriateness of used or used materials to be utilized. 2) Currently the process of determining the material feasibility of PT. Nusa Abadi Elektrika was carried out orally and saw the condition of the material with the naked eye and then recorded by the officer on a piece of paper which was later copied by the officer to the computer with the Microsoft Excel application. 3) With the creation of this decision support system, it is hoped that it can simplify the work process of warehouse officers in determining the feasibility of materials and minimize errors in managing material data.