Combining Bitemporal Conceptual Datamodel with Multiway Join Relations for Forecasting

Abstract

Abstract The Conceptual Modelling deals with representing an application domain in a descriptive and consistent manner without any computer metaphor. In this paper, we have modelled forecasting a temporal database adopting the concept of Bitemporal Conceptual Data Modelling considering the domain of Stock Exchange. The Novelty of this work is to optimize and execute data from multiple heterogeneous data sources by organizing the data through various transformations for preprocessing. We propose an algorithm called called enhanced 2P-TO (2 Phase Transformation Ordering) for efficient organization of non-overlapping sets of transformations. The salient feature of this approach lies in the consideration of direct, heterogeneous links among transformations and multiple data resources. Experiments are conducted using NASDAQ data which has 33280 tuples with 6 major attributes.. The efficiency of the algorithm is studied using two parameters: cardinality and number of correlated attributes. The correlation between the attributes is studied graphically using descriptive analysis. Forecasting of share trend is done for various attributes of the dataset and the best attribute for forecasting is found by adjusting its smoothing factors.