dc.description.abstract | The aim of this study was to develop and compare different forecasting models
for individual groups of perishable products and suggest the best forecasting
model for each group of perishable products. Through this study, it is intended
that businesses will be able to foresee fluctuations in demand for their perishable
products and therefore be able to make informed business decisions effectively.
In this study, statistical forecasting methods i.e., Moving Average, Weighted
Moving Average, Exponential Smoothing and Exponential Moving Average and
two modern machine learning models for forecasting i.e., XGBoost and ARIMA
were used. The data used was provided by a supermarket within Kampala and
was recorded from 1st/January/2013 to 15th/August/2021.
The raw data provided was cleaned using pandas and NumPy python
programming language libraries and the statistical analysis on the cleaned data
was performed in Microsoft Excel and the error measures used were Mean
Absolute Error (MAE), Mean Square Error (MSE), Mean Absolute Deviation
(MAD) and Mean Absolute Percentage Error (MAPE).
The results from the 3 perishable groups chosen for this study revealed that the
best forecasting model for the group of Meat was Moving Average with an
interval of 7 days. For the group of Diary, the suggested best forecasting method
is Weighted Moving Average with weights distributed as
0.3,0.2,0.15,0.13,0.12,0.07,0.03 for the selected 7 days and for the group of
Bread and Bakery, the suggested best forecasting method is Exponential Moving
Average (EMA) with a smoothing factor, Alpha=0.9.
On the other hand, Chapter 5 consists of the conclusions and recommendations
from this study. | en_US |