A joint and simultaneous prediction framework of weekday and weekend daily-activity travel pattern using conditional dependency networks

Daily activity pattern (DAP) prediction models within the Activity-based Modelling paradigm are being currently developed without adequate consideration of the various interdependencies among activities within a multi-day planning horizon. We hereby propose a conditional dependency network structure based interdependent multilabel-multiclass classification framework for joint and simultaneous prediction of weekday and weekend DAP of an individual. The prime advantage of the proposed modelling framework is flexibility of application of any algorithm for parameter estimation. Random Forest Decision Tree (RFDT), eXtreme Gradient Boosting and Light Gradient Boosting Machine (LightGBM) as the base classifier and probabilistic and non-probabilistic inference approaches are explored for measuring their comparative performance to provide insights for future researchers. Several variables representing neighbourhood characteristics are also investigated as DAP de-terminants along with socio-economic characteristics of individuals for the first time.This model is estimated based on two-days (weekday and weekend) activity-travel diary of 1808 households (6521 individuals) in Bidhanangar Municipal Corporation, India. The non-probabilistic approach-based models are found to achieve higher accuracy (0.81-0.92) compared to probabilistic models (0.76 to 0.82). RFDT and LightGBM are found to be the best performers in the probabilistic and non-probabilistic frameworks respectively. External validation results show that all proposed multiday-interdependent models (80%-94%) perform better than independent models (64%-83%).This framework can be applied to other transportations planning problems like household interaction in ac-tivity generation, joint destination and mode choice. This is also one of the first attempts to investigate the determinants of DAPs of urban commuters in an emerging country like India.

Determining dependence, centrality, and dynamic networks between green bonds and financial markets.

We adopted a network approach to examine the dependence between green bonds and financial markets. We first created a static dependency network for a given set of variables using partial correlations. Secondly, to evaluate the centrality of the variables, we illustrated with-in system connections in a minimum spanning tree (MST). Afterward, rolling-window estimations are applied in both dependency and centrality networks for indicating time variations. Using the data spanning January 3, 2011 to October 30, 2020, we found that green bonds and commodity index had positive dependence on other financial markets and are system-wide net contributors before and after COVID-19. Time-varying dynamics illustrated heightened system integration, particularly during the crisis periods. The centrality networks reiterated the leading role of green bonds and commodity index pre- and post-COVID. Finally, rolling window analysis ascertained system dependence, centrality, and dynamic networks between green bonds and financial markets where green bond sustained their positive dependence all over the sample period. Green bonds' persistent dependence and centrality enticed several implications for policymakers, regulators, investors, and financial market participants.