Forecasting in data pauperism

Abstract

Any forecasting method requires some evidences on the basis of which a future event can be estimated Sometimes these evidences are present as numerical quantities and the employable techniques are called quantitative forecasting techniques Sometimes the evidences cannot or should not be represented as numerical quantities instead they may better be represented as linguistic values In such cases quantitative forecasting techniques are not useful and a fuzzy analysis is the only solution A good quality of these fuzzy based forecasting techniques is that the constraints that are applicable over traditional quantitative forecasting techniques are relaxed in the case of fuzzy based techniques For example minimum requirement for evidences in the traditional quantitative forecasting techniques is 50 and more than 100 are preferable, whereas a fuzzy based forecasting technique is found to produce good forecasts with as low as 32 evidences In some cases these evidences may be missing or unavailable which is often the case with new or innovative products The forecasting techniques that are employable in such conditions are called qualitative forecasting techniques Such techniques make use of experts opinions and complex decision making in order to derive useful forecasts This involvement of human experts makes the process vulnerable to a number of biases and imprecision Again to deal with such imprecision to some extent a fuzzy assessment of the problem may be useful newlineThe thesis examines the literature on forecasting to find the techniques of merit when the evidences are either limited or not present at all The objective of the thesis is to find an answer to the poor forecasting accuracy obtained in case of the dearth of evidences newlineData pauperism in the context of this work refers to the two states of data under which forecasting are performed newlineAn absolute absence no prior data are available to perform any evidence based forecasting newline newline