Jämförelse av olika prognosticeringsmetoder - en fallstudie av Yaskawa Nordic AB
(2019) MIOM05Production Management
- Abstract
- Companies always strives to acheive as eective inventory control as possible.
At the same time, a suciently high delivery accuracy to satisfy the customer
is pursued. One prerequisite for successfully accomplishing this is to anticipate
customer demand in advance. An approximation of the demand in advance
is called forecasting. Forecasting implies that dierent techniques are used to
predict future demand. The techniques used can be based on quantitative data
or on qualitative data, even combinations of the corresponding methods occur.
A quantitative forecast is based on pure historical data, while a qualitative is
mainly established through experience. There are dierent areas of use for the
various techniques. This thesis covers... (More) - Companies always strives to acheive as eective inventory control as possible.
At the same time, a suciently high delivery accuracy to satisfy the customer
is pursued. One prerequisite for successfully accomplishing this is to anticipate
customer demand in advance. An approximation of the demand in advance
is called forecasting. Forecasting implies that dierent techniques are used to
predict future demand. The techniques used can be based on quantitative data
or on qualitative data, even combinations of the corresponding methods occur.
A quantitative forecast is based on pure historical data, while a qualitative is
mainly established through experience. There are dierent areas of use for the
various techniques. This thesis covers four such models. They are the naive
model, simple moving average, exponential smoothing and ARIMA.
Forecasts are mostly inaccurate. To cover the error, a safety stock is dimensioned.
The safety stock is much dependent on the distribution of the demand
during the lead time. The analysis of this thesis also covers the safety stock for
the case company. The distributions that are covered are: normal distribution,
negative binomial distribution, gamma distribution, poisson distribution and the
exponential distribution.
The purpose of the thesis is to produce a decision basis for the ordering of
robotics based on the historical demand data. The basis is going to support the
dierent forecasting models that are chosen. The suitable models are the ones
that entails the least deviation from the real data. The decision basis is also
going to contribute to optimisation of inventory, without any risk of impair the
delivery accuracy.
The case company that the empirical data is gathered from is Yaskawa Nordic
AB. They are a subsidiary to the main concern that is based in Japan. They
specialize in the robotic market and sell over 100 products. This thesis is only
going to analyse the 26 most frequent products.
The thesis research design is based on case study, with elements of modelling.
As mentioned above, there are four forecast models that are analysed. They
are modeled with the case company's empirical data. The thesis is based on a
deductive approach, where theories are studied and already existing models are
used.
After analysing the four dierent forecasting models for the 26 products, the
conclusion is that ARIMA is the model that was suitable for most products.
This model covered 54% of the products. The remaining 46 percent is divided between simple moving average and exponential smoothing, with six products each. The naive method is not suitable for any product. For the safety stock, the conclusion is that all the products are following an exponential distribution during the lead time. (Less)
Please use this url to cite or link to this publication:
http://lup.lub.lu.se/student-papers/record/8995719
- author
- Ramljak, Daniel and Ramljak, David
- supervisor
- organization
- course
- MIOM05
- year
- 2019
- type
- M1 - University Diploma
- subject
- other publication id
- 19/5625
- language
- Swedish
- id
- 8995719
- date added to LUP
- 2019-09-26 12:41:34
- date last changed
- 2019-09-26 12:41:34
@misc{8995719, abstract = {{Companies always strives to acheive as eective inventory control as possible. At the same time, a suciently high delivery accuracy to satisfy the customer is pursued. One prerequisite for successfully accomplishing this is to anticipate customer demand in advance. An approximation of the demand in advance is called forecasting. Forecasting implies that dierent techniques are used to predict future demand. The techniques used can be based on quantitative data or on qualitative data, even combinations of the corresponding methods occur. A quantitative forecast is based on pure historical data, while a qualitative is mainly established through experience. There are dierent areas of use for the various techniques. This thesis covers four such models. They are the naive model, simple moving average, exponential smoothing and ARIMA. Forecasts are mostly inaccurate. To cover the error, a safety stock is dimensioned. The safety stock is much dependent on the distribution of the demand during the lead time. The analysis of this thesis also covers the safety stock for the case company. The distributions that are covered are: normal distribution, negative binomial distribution, gamma distribution, poisson distribution and the exponential distribution. The purpose of the thesis is to produce a decision basis for the ordering of robotics based on the historical demand data. The basis is going to support the dierent forecasting models that are chosen. The suitable models are the ones that entails the least deviation from the real data. The decision basis is also going to contribute to optimisation of inventory, without any risk of impair the delivery accuracy. The case company that the empirical data is gathered from is Yaskawa Nordic AB. They are a subsidiary to the main concern that is based in Japan. They specialize in the robotic market and sell over 100 products. This thesis is only going to analyse the 26 most frequent products. The thesis research design is based on case study, with elements of modelling. As mentioned above, there are four forecast models that are analysed. They are modeled with the case company's empirical data. The thesis is based on a deductive approach, where theories are studied and already existing models are used. After analysing the four dierent forecasting models for the 26 products, the conclusion is that ARIMA is the model that was suitable for most products. This model covered 54% of the products. The remaining 46 percent is divided between simple moving average and exponential smoothing, with six products each. The naive method is not suitable for any product. For the safety stock, the conclusion is that all the products are following an exponential distribution during the lead time.}}, author = {{Ramljak, Daniel and Ramljak, David}}, language = {{swe}}, note = {{Student Paper}}, title = {{Jämförelse av olika prognosticeringsmetoder - en fallstudie av Yaskawa Nordic AB}}, year = {{2019}}, }