under the guidance of

Improving the accuracy of forecasting process is necessary to uplift the quality of man-agement decisions. From the earlier work in this area one can conclude that preprocessing (decomposition) and post-processing (combining) when applied to the time series result in in-crease in the accuracy of the forecasts which is the main aim of the forecasting process. In this work, we have concentrated on improving the rank-based combining technique proposed by [Set05]. We examine the two major factors that affect the performance of rank-based combining method like DTOPK- K-value selection and Ranking of experts at each point. We provide methods (K-diagram based method, Optimal Method) to get insight into these problems and based on our observations later develop a practical method of combining forecasts called ‘Modi-fied DTOPK ’ which gives substantial improvement in the accuracy of forecasts. We also cover in brief the three different forecasting methods that use one-point lookahead definition of trend in-stead of the standard one-sided trend definition. We report the results showing the performance of different combining techniques by running them over the forecasts given by these methods.