Planning Production Capacity Using Time Series Forecasting Method and Linier Programming

The objectives of this research is to determine the amount of production planning capacity sow talc products in the future utilizing previous data from january to december in year 2017. This researched considered three forecasting method, there are Weight Moving Average (WMA), Moving Average (MA), and Exponential Smoothing (ES). After calculating the methods, then measuring the error value using a control chart of 3 (three) of these methods. After find the best forecasting method, then do linear programming method to obtain the exact amount of production in further. Based on the data calculated, the method of Average Moving has a size of error value of Mean Absolute Percentage Error of 0.09 or 9%, Weight Moving Average has a size error of Mean Absolute Percentage Error of 0.09 or 9% and with Exponential Method Smoothing has an error value of Mean Absolute Percentage Error of 0.12 or 12%. Moving Average and Weight Moving Average have the same MAPE amount but Weight Moving Average has the smallest amount Mean Absolute Deviation compared to other method which is 262.497 kg. Based on the result, The Weight Moving Average method is the best method as reference for utilizing in demand forecasting next year, because it has the smallest error size and has a Tracking Signal not exceed the maximum or minimum control limit is ≤ 4. Moreover, after obtained Weight Moving Average method is the best method, then is determine value of planning production capacity in next year using linier programming method. Based on the linier programming calculation, the maximum amount of production in next year by considering the forecasting of raw materials, production volume, material composition, and production time obtained in one (1) working day is 11,217,379 pcs / year, or 934,781 pcs / month of finished product. This paper recommends the company to evaluate the demand forecasting in order to achieve higher business growth.


Introduce the Problem
Demand planning in general is known forecasting.Forecasting is an objective calculation using the data in the past, to specify something in the future.The necessary forecasting that must done by the company is product demand from customer requirements.By find out forecasting of production planning it can assist the company in determining the amount of product that should be in production.Therefore, with the forecasting, the company can achieve the goals and decision-making in determining the production capacity.However, in forecasting activities required the implementation of methods, it aims to minimize forecasting errors.
TEMPO Group is one of the Indonesian companies that engaged in an FMCG (Food Manufacturing Consumer Goods) and one of the business in this company is producing the powder for babies which need raw materials is talc.The company undertook its production demand planning still using conventionally method only on demand request, does not have a methods of Production Planning and Inventory Control, therefore, there is no effective method in order to calculate production planning forecasting.
Previous research states that only the implementation of forecasting methods, without calculating the planned production capacity.Therefore, the forecasting technique by calculating the maximum planning of production capacity is also intended to solve the problems that occur and perform an accurate production planning in response to customer demand in order to achieve effective and efficient value.

Explore Importance of the Problem
The results obtained advantage and contribute for learning development of science PPIC (Production Planning Inventory Control) and the company can be used as an input in activities of production planning.Moreover, production planning and control activities can be able controlled and can reduce wastage as well as maintaining consumer satisfaction.
Nasution and Prasetiawan (2008) described Demand forecast is the level of demand for products that are expected to be realized for a certain period of time in the future.Forecasting this request will be a very important input in the company's planning and control decisions.

Relevant Scholarship
Margaretha (2015) in their paper describes using time series forecasting techniques seems to be a good strategy if demand is highly volatile or resources are scarce.
Cheng, Akkarapol, Beyca, Le,Hui Yang, Zhenyu (2015) Time series forecasting has become essential toward advancing the way we manage and control complex real-world systems-from a conventional detectiondiagnosis-mitigation to a more proactive prediction-prognosis-prevention paradigm.

State Hypotheses and Correspondence to Research Design
The forecasting method is based on the use of variable relationship pattern analysis to be estimated with the time variable, which is the time series.Here are some of the Time Series methods used to find Forecast Demand (Demand Forecasting), such as: 1).Moving Average (MA) (Moving Average) is a method of forecasting that is act by taking a group of observation values, find out the average value as a forecast for the period in further.Single Moving Average method has special characteristics that is : a.To determine the forecast in the future period requires historical data over a period of time.For example, with a 3 month moving average, the 5th month forecast is only made after the 4th month finishes / ends.If the month of the 7th month moving average can only be made after the 6th month ends.b.Increasingly time period of moving average, the slicking effect is more visible in the forecast or resulting in a smoother moving average 2).Weight Moving Average (WMA) Weighted Moving Avarage Method is the same simple moving average calculation method, but it needs a coefficient of weighing and is used if there is a trend in the past pattern.Included in the quantitative method is the econometric model, the input-output analysis model and the regression model.The weighted moving average model is more responsive to change, since data from the new period is usually given greater weight.View the historical pattern of the actual data of the request, in the event of a substantial change or unstable movement during that time period.Use a weighted moving average forecasting model 3).Exponential Smoothing (ES) Exponential Smoothing method is the procedure of continuous improvement for forecasting for the latest observation object.This forecasting method emphasizes the exponentially decreasing priority of older observational objects.On the exponential smoothing there is one or more of the specified explicity smoothing parameter, and these results determine the weight imposed on the value of observation.

Method
The process of making this research through several stages which are interrelated and support each other.Starting with the preliminary studies by studying the previous studies used as a reference for determining the title to be appointed.After the title was set, next step is the creation of constraints and formulate the issues that will be discussed.After that, the creation of constraints and formulate the problem is identify to the need for data to be used and needed, which correspond to the theme and title and can meet the answers to the existing problems.In collecting data done by interview and literature study, interview conducted to the main person who in charge in production department.

Research Method
This research is included in quantitative research, which is about research that is calculating and systematic.In the process of quantitative research is a lot of doing calculations done systematically, well planned and structured.In the writing of this research, research methods conducted by determining the appropriate forecasting methods and can be applied in the company and also make the forescasting prediction in further.how to use the data processing as follows:

1). Tabulation
This technique arranges data in the form of tables or diagrams.In this research the table is used for forecasting process, error value and control chart 2).Data Analysis As to be obtained in processing the data are: a. Calculates demand forecasting using Moving Average, Weight Moving Average and Exponential Smoothing methods.
b. Comparing the results of forecasting obtained by using the error value and map control.c.Calculate production planning using linear simplex programming.

Data Actual
Forecasting is an important thing that must be conducted by the company in order to plan the needs of raw materials, therefore, it required to choice an appropriate forecasting methods in order the forecasting value results not big different from the reality.The demand data of talc powder raw material in year 2017 is shown in Figure 1 below.

Forecasting Method Calculation
In making this research the author uses the method of Time Series which include: Moving Average method, Weight Moving Average, and Exponential Smoothing.Then the calculation will be conducted as follows :

1). Moving Average (MA) method
Moving Average method is a method of forecasting where the number of data requests in the previous period is divided by the number of data in that period.After calculation in the control chart can be seen from the graph exponential smoothing method, the results of this method exceeds the maximum limit of the highest Signal Tracking ≥ 4 was in June that touched the figure of 4.08, so this method is less accurate demand planning activities.

Discussion
Based on the forecasting method calculation in above, can be seen from the error value (error) which is the smallest, means resulted from the error value can be known about the reason the risk of errors occurring.
Because being an forecaster person should have a mistake, and as much as a forecaster person have to known many mistakes but a forecaster cannot eliminate uncertainty, otherwise, only can reduce the uncertainty.Forecasting accuracy indicators are Mean Absolute Deviation (MAD), Mean Square Error (MSE) and Mean Absolute Percentage Error (MAPE).To know the control chart that is using Tracking Signal.After it is known which is the best method, and then subsequently do the determination of capacity planning the warehouse using the method of Simplex solver.Because the best method is already known then only the best method is shown in the map control.The Assessment undertaken include: 1).Priority 1 = Total production of powder per shift of at least 2 (may be more, not less) 2).Priority 2 = Production of 1 talc required 800 gr talc plus 500 gr of mixed material 3).Priority 3 = Production time is equal to (approximate) time available (should not be more, may be less) In accordance with a linear equation then obtained the equation: By using the simplex program solver, then obtained a production target per year as follows: Based on the calculation, the maximum amount of production will be obtained by considering the forecasting of raw materials, production, material composition, and production time in 1 day by 11,217,379 pcs / year, or 934,781 / month of finished powder products.

Conclusion
The method that obtained from Time Series forecasting method calculation, there are Moving Average which has error value from MAD 284.598,5 pcs & MAPE 0.09, then Weight Moving Average method has MAD value from 262,496,6 pcs & MAPE equal to 0.09, and Exponential Smoothing method has the error value of MAD is 370.712,5pcs & MAPE of 0.12.Weight Moving Average method has MAPE value below 10%, so it has a very good performance in demand forecasting.To know the average of absolute error the weight moving average method also has the smallest Mean Absolute Deviation (MAD) value so that the result of forecasting will be more accurate.Production planning using the selected time series method is the weight moving average method

D
of time series used D't = forecasting for the coming period Dt = demand actual in period t

Figure 3 .
Figure 3. Tracking signal moving average method

Fig
Fig

Table 2 .
Weight coefficient of weight moving average

Table 5 .
Results of forecasting methodAfter the comparison, there are 3 elements that can determine the accuracy of forecasting.Mean Absolute Percentage Error is a measure of the relative error.MAPE is usually more means than MAD because the percentage error MAPE stating the results of the actual demand against the forecast during a certain period which would provide information the percentage error is too high or too low.The above shows that the Moving Average and the Moving Average has a Weight smaller than MAPE Exponential Smoothing that is has the same value of 0.09 or 9%.However, the Weight Moving Average has a size smaller than the MAD Moving Average, so the Weight Moving Average Method is a method which is best among the Moving Average and Exponential Smoothing and can be used as recommendations for PT TEMPO Group in conducting product demand forecasting.

Table 6 .
Calculation of data