 Open Access
 Total Downloads : 434
 Authors : Pradeep Ku Sahu, Rajesh Kumar
 Paper ID : IJERTV2IS90048
 Volume & Issue : Volume 02, Issue 09 (September 2013)
 Published (First Online): 04092013
 ISSN (Online) : 22780181
 Publisher Name : IJERT
 License: This work is licensed under a Creative Commons Attribution 4.0 International License
The Evaluation of Forecasting Methods for Sales of Salted Butter Milk in Chhattisgarh, India
Pradeep Ku Sahu, Rajesh Kumar
Chhatrapati shivaji Institute of Technology, Durg, Chhattisgarh, India
Abstract
The purpose of in this paper is to identify most appropriate forecasting method for sales of salted butter milk in Chhattisgarh. Applying weekly data spreading over October 2011 to October 2012, on the sales of salted butter milk in liter. The forecasting method analyzed included: nave model , moving average, double moving average, simple exponential smoothing; and semi average method. The accuracy of the forecasting method was measured using mean Forecast Error (MFE), mean Absolute Deviation (MAD), mean Square Error (MSE), root mean square Error (RMSE).

Introduction
Forecasting is a prediction of what will occur in the future. It is an uncertain process that is vital to survival in todays international business environment. Uncertainty demand of the product increases or decreases the production cost, customer satisfaction & employee moral etc. Managers try to forecast with as much accuracy as possible.
Forecasting is a critical component of Supply Chain Management. The supply chain involves everything that pertains to producing a product or service from a companys suppliers all the way to the customers. Forecasting help determine the amount of inventory to be kept on hand, how much product should be made. Inaccurate forecasting can lead to costly inventory buildup or stock outs. Both of these events are harmful in a business world where customer service is of almost importance.

Components of Demand Forecasting

Time Frame

Behavior and the possible existence of patters Time Frame
The length of forecasting depends on product market changes and susceptibility to technological changes.
The classifications are generalizations. Short, Mid and Long range is all relative to the business and what is being forecast.
Short to Mid Range forecast are usually anywhere form daily to upto two years in legth. They are commonly used to determine production and delivery schedules and to establish inventory levels.
Long range forecasts are generally over two years into the future. They are usually used for strategic planning. Strategic planning determines where the company is headed in the future. It is used to establish long term goals plan new products enter new markets and develop new facilities & technology.
Behavior
Demand sometimes behaves in random and irregular ways. Other times it exhibits predictable behavior. The three main types of predictable behavior are trend, cycles and seasonal patterns.
A trend is a gradual, longterm, upward or downward movement in demand. A current trend is the steady increase in sales of personal computers over the past few years.
A cycle is an up and down movement in demand that repeats itself over a longer time span. Automotive sales often behave in cyclical pattern.
A seasonal pattern is a repetitive movement in demand that occurs periodically. Sales of winter sports equipment are seasonal by nature.



Literature survey
There has been a great deal of discussion in economic literature about applications of various forecasting models for forecasting desired issues. Several time series forecasting techniques such as nave model, moving average, double moving average, simple exponential smoothing; and semi average method has been applied to forecasting. In a study, Cacatto et al. (2012) introduced the forecasting practices that have been used by food industries in Brazil and detected how these companies use forecasting methods. The
data were analyzed by multivariate statistics techniques using the SPSS software. They stated that the companies did not use sophisticated forecasting methods, the historical analysis model was the mostly used. In an attempt Ryu et al. (2003) evaluated the forecasting method for institutional food service facility. They are identifying the most appropriate forecasting method of forecasting meal count for an institutional food service facility. The forecasting method analyzed included: nave model 1,2 and 3; moving average method, double moving method, exponential smoothing method, double exponential method, Holts method, winter method, linear regression and multiple regression method. The accuracy of forecasting methods was measured using mean absolute deviation, mean squared error, mean percentage error, mean absolute percentage error method, root mean squared error and Theils Ustatistic. The result of this study showed that multiple regressions was the most accurate forecasting method, but nave method 2 was selected as the most appropriate forecasting method because of its simplicity and high level of accuracy. While Strasheim et al., (1992) introduced variety of alternative forecasting technique were evaluated for purpose of stock replenishment is an important function of part in the typical reordering motor vehicle spare parts with aim of selecting one optimal technique to be implemented in automatic reordering module of real time computerized inventory management system. A large number of forecasting technique were evaluated, namely simple moving average(Averages, Moving Averages, Double Moving Averages), Exponential Smoothing(Single Exponential Smoothing, Adaptive Exponential Smoothing, Double Exponential Smoothing (Brown's one 'parameter linear method and Holt's two parameter method), Triple Exponential Smoothing (Brown's one parameter quadratic method and Winter's three parameter trend and seasonality method)), linear Regression, Multiple Regression. The accuracy of forecasting methods was measured using statistical measures (mean error, mean absolute deviation, mean squared error), relative measures (percentage error, mean percentage error, mean absolute percentage error method) and others measures (Theils U statistic, Durban Watson value and forecasting

Data Collection
The data for this study were collected and recorded on weekly basis. The data contain sales of salted butter
milk from October 2011 to October 2012. All the data was saved into an Excel spreadsheet.
Week No.
Demand
Week No.
Demand
1
740.4
29
10804.2
2
722.4
30
8869.4
3
758.4
31
9835.8
4
465
32
9184.4
5
713.2
33
11352
6
544
34
12007.2
7
467
35
12366.8
8
469.2
36
13497.6
9
479
37
9156.6
10
353
38
1855
11
377
39
1715.8
12
292
40
1372
13
286
41
668
14
268
42
645
15
/td>
245
43
406.6
16
322.6
44
263
17
411
45
372.2
18
293.6
46
444
19
681.8
47
473
20
579.4
48
630.4
21
1278
49
588.2
22
2629
50
607
23
2901.8
51
607.6
Week No.
Demand
Week No.
Demand
1
740.4
29
10804.2
2
722.4
30
8869.4
3
758.4
31
9835.8
4
465
32
9184.4
5
713.2
33
11352
6
544
34
12007.2
7
467
35
12366.8
8
469.2
36
13497.6
9
479
37
9156.6
10
353
38
1855
11
377
39
1715.8
12
292
40
1372
13
286
41
668
14
268
42
645
15
245
43
406.6
16
322.6
44
263
17
411
45
372.2
18
293.6
46
444
19
681.8
47
473
20
579.4
48
630.4
21
1278
49
588.2
22
2629
50
607
23
2901.8
51
607.6
Table 1: Demand of salted butter milk (in liter)
24
4681
52
778.6
25
6914
53
788
26
7630
54
1014.4
27
8894
55
819.6
28
8054
56
588.6
Fig 1: Demand of salted butter milk (in liter)
Ft+1 = (Yt +Yt1 +Yt2 + +Ytn+1)/n Where:
Ft+1 = the forecast value for the next period Yt = the actual value at period t
n = the number of term in the moving average
The optimal n value can be determine by interactive model that the smallest error. In some method the general approach has been to use MSE. In this study, the value of n taking 1, 2 and 3.
4.1.2 Double Moving Average Method.
Hanke and Reitsch (1998) recommended the use of the double moving average method to forecast time series data. Forecasting with a double moving average requires determining two averages. The first moving average is computed; a second moving average is calculated. Five equations are used in the double moving average:
Mt = Ft+1 = (Yt +Yt1 +Yt2 + +Ytn+1)/n Mt = (Mt + Mt1 +Mt2 + +Mtn+1)/n
At = 2Mt – Mt

Methodology
This study evaluated different forecasting model using
Bt =
2
n 1
( Mt – Mt )
salted butter milk demand data from Raipur dugdh sangh(Devbhog) at Raipur (Chhattisgarh).Weekly data from October 2011 to October 2012 were collected and used to forecast the salted butter milk demand. The forecast model used in the analysis included simple moving average method, double moving average method, single exponential method (=0.1, =0.2, =0.3), semi average method, nave Model. The most appropriate forecasting method was determined on the basis of accuracy. In this research, several common accuracy methods were used: mean forecast error (MFE), mean absolute deviation (MAD), mean square error (MSE) and root mean square error (RMSE). The ranking was assigned to each forecasting method.

Forecasting Methods
4.1.1 Moving Average Method.
Ft+p = At + Bt p Where:
n = the number of period in the double moving average
Yt = the actual series value at time period t
P = the number of period ahead to be forecast

Simple Exponential Smoothing Method.
The exponential smoothing method is a technique that uses a weighted moving average of past data as the basis for a forecast. This method keeps a running average of demand and adjusts it for each period in proportion to the difference between the latest actual demand figure and the latest value of the average. The equation for the simple exponential smoothing method is:
The moving average method involves calculating the average of observations and then employing that average as the predictor for the next period. The moving average method is highly dependent on n, the
Ft+1
Where:
= Yt + (1) F
t1
number of terms selected for constructing the average. The equation is as follows:
Ft+1 = the new smoothing value or the forecast value for the next period
= the smoothing constant (0 < <1)
Yt = the new observation or actual value of the series in period t
Ft = the old smoothed value or forecast for period t
The accuracy of the simple exponential smoothing method strongly depended on the optimal value of ().The preferred range for is from 0.1 to 0.3. In this study, the value of taking 0.1, 0.2 and 0.3.

Semi Average Method.
According to this method, the original data are divided into two equal parts and the values of each part are then summed up and averaged. The average of each part is centered in the period of the time of the part from which it has been calculated and then plotted on graph. Then a straight line is drawn to pass through the plotted points. This line constitutes the semi average trend line. When the number of year is odd, the middle year is not considered while dividing the data into two equal parts and obtaining the average.

Nave Method.
Nave method are forecasting techniques obtained with a minimal amount of effort and data manipulation and are based on the most recent information available (Shim, 2000). The nave method uses data from the previous week to forecast the current week (one week lag):
of the measure being used, the lowest value generated indicates the most accurate forecasting model.

Mean Forecast Error.
Mean forecast error (MFE) is the mean of the deviation of the forecast demands from the actual demands.
n
n
(Yt Ft )
MFE = t 1
n
Where:
Yt = the actual value in time period t
Ft = the forecast value in time period t n = the number of periods

Mean Absolute Deviation.
A common method for measuring overall forecast error is the mea absolute deviation (MAD). Heizer and Render (2001) noted that this value is computed by dividing the sum of the absolute values of the individual forecast error by the sample size (the number of forecast periods). The equation is:
n
n
(Yt Ft )
Ft+1
Where:
= Yt
Where:
MAD = t 1
n
Ft+1 = the forecast value for the next period Yt = the actual value at the next period

Measuring Forecasting Error
There is no consensus among researcher as to which measure is best for determining the most appropriate forecasting method (Levine et al., 1999). Accuracy is the criterion that determines the best forecasting method; thus, accuracy is the most important concern in evaluating the quality of a forecast. The goal of the forecasts is to minimize error. Forecast error is the
difference between an actual value and its forecast
Yt = the actual value in time period t
Ft = the forecast value in time period t n = the number of periods

Mean Square Error.
Jarrett (1991) stated that the mean square error (MSE) is a generally accepted technique for evaluating exponential smoothing and other methods. The equation is:
1 n 2
value (Hanke & Reitsch, 1998).
Some of the common indicators used evaluate accuracy are mean forecast error, mean absolute deviation, mean
Where:
MSE =
(Yt Ft )
n
n
t 1
squared error, and root mean squared error. Regardless
Yt = the actual value in time period t
Simple Moving Average Method (n=3)
0.24405
981.43
3908031
1976.874
Simple Moving Average Method (n=4)
5.84643
1153.2
5019452
2240.41
Double Moving Average Method (n=2)
5.4786
667.05
2140258
1462.96
Double Moving Average Method (n=3)
3.2742
843.37
3362785
1833.79
Double Moving Average Method (n=4)
1.2567
1097.4
5104754
2259.37
Single Exponential Method(=0
.1)
174.100
2564.5
12039667.7
3469.822
Single Exponential Method(=0
.2)
10.5674
1830.6
7660178.71
2767.702
Single Exponential Method(=0
.3)
0.2531
1379.9
5354692.477
2314.020
Simple Moving Average Method (n=3)
0.24405
981.43
3908031
1976.874
Simple Moving Average Method (n=4)
5.84643
1153.2
5019452
2240.41
Double Moving Average Method (n=2)
5.4786
667.05
2140258
1462.96
Double Moving Average Method (n=3)
3.2742
843.37
3362785
1833.79
Double Moving Average Method (n=4)
1.2567
1097.4
5104754
2259.37
Single Exponential Method(=0
.1)
174.100
2564.5
12039667.7
3469.822
Single Exponential Method(=0
.2)
10.5674
1830.6
7660178.71
2767.702
Single Exponential Method(=0
.3)
0.2531
1379.9
5354692.477
2314.020
Ft = the forecast value in time period t n = the number of periods

Root Mean Square Error.

Root mean square error (RMSE) is the square root of MSE. This measures error in term of units that are equal to the original value (Jarrett, 1991).Symbolically, the equation is:
1 n
Where:
RMSE =
(Yt Ft )
n
n
t 1
Yt = the actual value in time period t
Ft = the forecast value in time period t n = the number of periods



Evaluation of Forecasting Method
In this study, the most appropriate forecasting method was selected on the basis of both level of accuracy and ease of use. The various forecasting method are using to forecast future demand of salted butter milk in Chhattisgarh, the accuracy of the forecasting method was assessed using mean forecast error (MFE), mean absolute deviation (MAD), mean square error (MSE), and root mean square error (RMSE).
In the case of forecasting of salted butter milk demand in Chhattisgarh, special consideration as to each methods ease of use was required, since the person in charge of forecasting usually has little time andin some instances little knowledge of how implement the forecasts.
Table 2: Summary of Forecast Accuracy (salted butter milk)
METHOD
MFE
MAD
MSE
RMSE
Simple Moving Average Method (n=2)
1.6821
765.82
2823812
1680.42
Semi average Method
126.907
3106.7
16780524.3
4096.40
Nave Model
2.710
646.01
1893561.34
1376.067

Result and Discussion
In this study, four accuracy model mean forecast error (MFE), mean absolute deviation (MAD), mean square error (MSE), and root mean square error (RMSE)were adopted to assess the accuracy of forecasting methods. The smaller the forecast error, the more accurate forecasting method.
Method
MFE
MAD
MSE
RMSE
Ranking Total
Overall Ranking
Simple Moving Average Method (n=2)
4
3
3
3
13
2
Simple Moving Average Method (n=3)
1
5
5
5
16
3
Simple Moving Average Method (n=4)
8
7
6
6
27
7
Double Moving Average
7
2
2
2
13
2
Method
MFE
MAD
MSE
RMSE
Ranking Total
Overall Ranking
Simple Moving Average Method (n=2)
4
3
3
3
13
2
Simple Moving Average Method (n=3)
1
5
5
5
16
3
Simple Moving Average Method (n=4)
8
7
6
6
27
7
Double Moving Average
7
2
2
2
13
2
Table 3: Overall Ranking of Forecasting Method for salted butter milk
Method (n=2)
Double Moving Average Method (n=3)
6
4
4
4
18
4
Double Moving Average Method (n=4)
3
6
7
7
23
5
Single Exponenti al Method(
=0.1)
11
10
10
10
41
9
Single Exponential Method(= 0.2)
9
9
9
9
36
8
Single Exponenti
al Method(
=0.3)
2
8
8
8
26
6
Semi average Method
10
11
11
11
43
10
Nave Model
5
1
1
1
8
1
Nave Method using the last week of data to forecast the next week. It has the lag of one week. Nave
Method had small error (MFE = 2.710, MAD =
646.010, MSE = 1893561.34, RMSE = 1376.067)
outperformed all the other methods.
Simple Moving Average Method (n=2) & Double Moving Average Method (n=2) was ranked second because it had small errors and the total ranking of the semi average method is 13 as shown in Table 3. So this method ranked is second.
Simple Moving Average Method (n=3) was ranked second because the total ranking of this method is 16 as shown in Table 3. When taking n value 3, single exponential method (SEM) obtained the third minimum errors (MFE= 0.24405, MAD = 981.4321, MSE =
3908031, RMSE = 1976.874) as shown in Table 2.
Double Moving Average Method (n=3) was ranked fourth because both total ranking is 18 as shown in Table 3. In simple moving average method with n=3 produced fourth smallest error as shown in Table 2.
Double Moving Average Method (n=4) produced large errors (MFE= 1.2567, MAD = 1097.47, MSE =
5104754, RMSE = 2259.37) as compare to Nave Method, Simple Moving Average Method (n=2, 3) & Double Moving Average Method (n=3). So this method ranked is fifth.
Single Exponential Method (=0.3) was ranked sixth because it had large errors (MFE = 0.2531, MAD = 1379.998, MSE = 5354692.477, RMSE = 2314.0208)
and total ranking is 26 as shown in Table 3.
Simple Moving Average Method (n=4) was ranked seventh because both total ranking is 27 as shown in Table 3. In Simple Moving Average Method with n=4 produced seventh smallest error as shown in Table 2.
Single Exponential Method (=0.2) was ranked eight because the total ranking of this method is 36 as shown in Table 3. Single Exponential Method (=0.2) obtained the eight minimum errors (MFE= 10.5674, MAD = 1830.67, MSE = 7660178.71, RMSE =
2767.7027) as shown in Table 2.
Single Exponential Method (=0.1) was ranked ninth because total ranking is 41 as shown in Table 3. In Double Moving Average Method with =0.1produced ninth smallest error as shown in Table 2.
Semi average Method had large error (MFE = 126.907, MAD = 3106.73, MSE = 16780524.3, RMSE =
4096.40) as shown in Table 2. So this model ranked is tenth.

Conclusions
This study identified the most appropriate forecasting method based on accuracy and simplicity. The result showed that Nave Method obtained the best accuracy; however, it was selected as the most appropriate forecasting method for sales forecasting of salted butter milk in Chhattisgarh.
References

Heshamk k. Alfares and Mohammad Nazeerudin (2002) Electric load forecasting: literature survey and classification method, International journal of system science: volume 33, Number 1, pp 2324.

Cristiano Cacatto, Patricia Belfiore and Jose Geraldo Vidal Vieira (2012) Forecasting Practices in Brazilian food Industry, Journal of logistics management: Vol. 1, No. 4, pp 2436

Ryu, Kisang and Sanchez, Alfonso (2003) The Evaluation of Forecasting Method at an Institutional Foodservice Dining Facility, Journal of Hospitality Financial Management: Vol. 11: Iss.1, Article 4.

Rachel J.C. Chen, Peter Bloomfield and John S. Fu (2003) An Evaluation of Alternative Forecasting Method to Recreation Visitation, Journal of Leisure Research: Vol. 35, No.4, pp 441454.

Strasheim et al., (1992) Demand Forecasting for Motor Vehicle Spare Parts, A Journal of Industrial Engineering, Vol. 6, No. 2, pp 1819.

Mihaela Bratu et al The Accuracy of Unemployment Rate Forecasts In Romania and The Actual Economic Crisis, Scientific Bulletin Economic Sciences, Vol.11, Issue 2, pp 5667.

Feridun, M. & Adebiyi,M.A.(2006) Forecasting Inflation in Developing Economics: The Case of Nigeria, International Journal of Applied Econometrics and Quantitative Studies, 19861998, Volume 3, Issue 1, pp 18 19.

Floros, Ch., et al., (2005) Forecasting the UK Unemployment Rate: Model Comparisons, International Journal of Applied Econometrics and Quantitative Studies, Vol.2, Issue 4, pp 5772.

Sharma, A.K., Gupta, A and Sharma, U., (2013) Electricity forecasting of Jammu & Kashmir: A Methodology Comparision,International Journal of
Electrical Engineering & Technology (IJEET), Vol.4, Issue 2, pp 416426.

Rachel J.C. Chen, Peter Bloomfield and Frederick W. Cubbage (2008) Comparing Forecasting Models in Tourism, Journal of Hospitality & tourism Research, Vol.32, Issue 1, pp 321.

Patil, D.P., Shrotri, A.P. and Dandekar, A.R., (2013) Management of Uncertainty in Supply Chain, International Journal of Emerging Technology and Advanced Engineering, Vol.2, Issue 5, pp 303308.

Carol T. West, et al., (2003) The Status of Evaluating Accuracy of Regional Forecasts, The Review of Regional Studies, Vol.33, Issue 1, pp 85103.

Paul, S.K., et al., (2011) Determination of Exponential Smoothing Constant to Minimize Mean Square Error and Mean Absolute Deviation, Global Journal of Research in Engineering, Vol.11, Issue 3, Version 1.0.

Lim, P.Y., and Nayar, C.V., (2012) Solar Irradiance and Load Demand Forecasting Based on Single Exponential Smoothing Method, IACSIT International Journal of Engineering and Technology, Vol.4, Issue 4, pp 451455.

Armstrong, J.S. and Collopy, F., (1992) Error Measures For Generalizing About Forecasting Method: Empirical Comparisons, International Journal of orecasting, Vol.8, pp 6980.

Padhan, P.C., et al., (2012) Use of Univariate Time Series Model For Forecasting Cement Production in India, International Research Journal of Finance and Economics, Issue 83.

Panneerselvam, R., et al., (2009) Production and Operation Management, 2nd edition, PHI Learning Private Limited, New Delhi (India).

Chopra, S. and Meindl, P., (2010) Supply Chain Management Strategy, Planning and Operation, Pearson, 4th Ed, India.

Ramamurthy, P., et al., (2005) Production and Operation Management, New Age International (P) Limited, Publishers, New Delhi (India).

Chary, S.N., et al.,(2009) Production and Operation Management, Tata McGrawHill, New Delhi (India)

Sahu, P. and Kumar, R.(2013) Survey of demand fluctuation of dairy products at Raipur Dugdh Sangh,
Chhattisgarh, India for selection of appropriate forecasting Uncertainty Supply Chain Management, Vol.1, Issue 3, pp 133144.

Sahu, P. and Kumar, R.(2013) Demand Forecasting For Sales of Milk product (Paneer) In Chhattisgarh International Journal of Inventive Engineering and Sciences, Vol.1, Issue 9,pp1013.