Forecasting is a method of estimation of the unknown parameters and it is being mainly used in the time series data.
In supply chain management demand prediction and forecasting is one of the main activity within an organization. The importance of the demand forecasting in a supply chain is for high customer satisfaction and inventory management system. The forecasting helps to predict product demand so that it can fulfill and satisfy the customer needs with short lead time or Just in Time (JIT).
On the other hand, it optimizes the inventory cost leads to the organizational profit. It also helps the better product planning and marketing strategies. 1.2 Objective: The main objective of this project is to forecast the demand of the product using the various well-known methods. We will compare the forecasting methods and evaluate the performance of the methods. We have a set of sample data of an electronics manufacturing company.
Using those data set we compare the forecasting methods and understand how an error can be minimized in the forecasting procedure so that it can be a value added service to our customers. 1.3 Scope of the work: There are varieties of activity involved in the background before forecasting the demand of the product. Data collection is one of the major activities and continuous process in the forecasting methods. Here we have sample sets of historical data of a electronics manufacturing company but the real and continuous data set is out of scope for this project. We can collect the varieties of data set from the open data source like Kaggle (https://www.kaggle.
com/), Data Science Central (http://www.datasciencecentral.com/) to evaluate the models.The primary focus of this project is to evaluate and fit an optimise model in the currently available data set. Further, we will try to fit in the model from the other types of the data set. 2 Literature Review:2.1 Review Paper – “Stan: A probabilistic programming language” Work on probabilistic programming language states that Stan has many related advantages. Stan provides the interface as “cmdstan” for the command line shell, “pystan” for Python and “rstan” for R language.
It also provides wrapper packages for MATLAB, Julia, Stata, and Mathematica. It supports the operating system like Windows, Mac OS X, and Linux, and are open-source licensed. (Carpenter et al, 2016)Remarks: The package and methodology can help to implement the forecasting methods and models.2.2 Review Paper – “Locally weighted regression: an approach to regression analysis by local fitting” The research paper by Cleveland and Devlin describes on Locally weighted regression (loess) method. This is a multivariate smoothing procedure computed based on moving average method for a time series. (Cleveland et al, 1988)Remarks: This paper will help to forecast the demand of a product. The concerned research paper is well-known for the least-squares fitting of parametric functions and the method “loess” can be applied to the time series data set.
2.3 Review Paper – “25 years of time series forecasting” The next paper that I seek to analyze is by Gooijer and Hyndman. This paper is a review of all the writings on forecasting published in Journal of Forecasting (1982–1985) and International Journal of Forecasting (1985–2005). It analyses the methods on time series forecasting. (Gooijer et al, 2006)Remarks: The concerned research paper will help me to adopt the relevant methods in my project work. The relevant methods are -· Exponential smoothing· Auto-Regressive Integrated Moving Average (ARIMA).2.4 Review Paper – “Automatic time series for forecasting: the forecast package for R” The research paper by Hyndman and Khandakar describes and implements a “forecast” package for R.
It concentrates on mainly two methods, which are Exponential smoothing and Auto-Regressive Integrated Moving Average (ARIMA). (Hyndman et al, 2007)Remarks: The package and methodology discussed in this paper can help me to implement the forecasting methods and models. 3.1 Problem Statement: The present research is focused on analyzing the relationship between production and prediction. 3.2 How you propose to carry out the project: · Collect the historical data from the organization and other open sources.· Analyse the data and compare the methods of forecasting as described in the above mentioned research papers.
· Though there are several methods which enable one to forecast however my goal will be to arrive at a conclusive method which will be conducive to least error. 3.3 MethodsThe methods that I will adopt in my project work are -· Time series methods of forecasting.o Additive Forecasting model, which will be the component of Growth function, Seasonality and aperiodic (like holidays) function.
; Where is a growth function, is a periodic or seasonality function, is aperiodic function and is a random error. · Moving Average.· Exponential Smoothing.· ARIMA (Auto-regressive Integrated Moving Average)