Forecasting Sales Using Neural Networks PDF
Forecasting Sales Using Neural Networks PDF
Forecasting Sales Using Neural Networks PDF
turnover x t +1
3000 3000 difference
.
2500 2500
t time
.
1 T
2000 2000
0 0
36 41 46 51 4 9 14 19 24 29 34 39 44 49 2 7 12
holidays
x
: : resp.
networks with one hidden layer together with the back-
propagation training method. In order to predict the zt = t; +05
x
c
:
given in the input layer. The only result in the output where min and max are the minimum and maximum
layer is the sale for the next week. values of time series and and are the average
x
Due to the purchasing system used in the supermar- and the standard deviation. is a factor to control the c
ket there is a gap of one week between the newest sale interval of the values.
value and the forecasted week. In addition the pric- For the prices the most eecting indicator is the
ing information, advertising campaigns and holidays price change. So the prices are modeled as follows:
are already known for the future, when the forecast is 8 0 9 : price increases 9
calculated. This information is also given to the input < =
:= : 0 0 : price keeps equal within
:
3. Preprocessing the Input Data For both the time series of holidays and advertising
campaigns we tested binary coding and linear aggrega-
An ecient preprocessing of the data is necessary tion to make them weekly. Their indicators are:
to input it into the net. In general it is better to 8 0 9 : if there is a holiday resp. 9
transform the raw time series data into indicators that < : =
represent the underlying information more explicitly. yt := : advertising within week resp. t
the back-propagation learning algorithm 4]. We tested (normalized number of special days within week ) t
ing method is often used by the supermarket's person- Based on the information in table 1 the naive ap-
nel. proach is outperformed by the two other methods. For
18 of the 20 items the prediction by the neural network
4.2. Statistical Prediction is better than the statistical prediction method.
A close inspection of the times series favored by the
The statistical method is currently being used by statistical approach shows that these are very noisy
the supermarket's headquarters to forecast sales and without any implicit rules that could be learned by the
to guide personnel responsible for purchasing. It calcu- neural network. Especially one of these items has an
lates the moving average of a maximum of nine recent average weekly sale of less than 4 items.
weeks, after these sale values have been ltered from Figure 3 shows the predicted values for item 468978
exceptions and smoothed. calculated by the statistical and neural approach. The
price, advertising and holiday information is included
4.3. Neural Prediction in this gure.
We reached good results for = 2 recent values of
n
the sale time series in the sliding window. The other 5. Conclusions and Future Research
inputs are, one neuron each: both the dierence of the
sale ( 0t = t ; t;1) and the turnover of the whole
x x x For a special group of items in a German super-
group of items for the last week and the holiday, ad- market neural nets have been trained to forecast fu-
vertising and pricing information for the week to be ture demands on the basis of the past data augmented
predicted. with further inuences like price changing, advertising
Thus, for each item a net with 7 input neurons and 4 campaigns and holiday season information. The ex-
hidden neurons is trained for a one week ahead forecast perimental results show that neural nets outperform
with a gap of one week. We reached better results with the naive and statistical approaches that are currently
the binary scaling for the holiday and advertising time being used in the supermarket.
weeks 13/1995 to 13/1996 legend
Price 468978
6.00
45
5.00 40
35 Sale Prediction 468978
Statistical
4.00 30
Sale Prediction 468978
25 Neural Network
3.00
20
2.00 15
10
1.00
5
0.00 0
13 17 21 25 29 33 37 41 45 49 1 5 9 13
Holidays
Figure 3. Comparison of sale prediction for an item by statistical and neural approach