Time series causal relationships discovery through feature importance and ensemble models

Abstract

Inferring causal relationships from observational data is a key challenge when seeking to understand the interpretability of Machine Learning models. Given the ever increasing amount of observational data available in many areas, Machine Learning algorithms used for forecasting have increased their complexity, leading to a less understandable path of how a decision is made by the model. With this in mind, we propose leveraging ensemble models, e.g., Random Forest, to assess which input features the trained model pays more attention to when making a forecast and, in this way, establish causal relationships between the variables. The advantage of those algorithms is the possibility of obtaining the feature importance, which allows us to build the causal network. We apply our methods to estimate causality in time series for two domains: One climate dataset and two oil field production datasets. We aim to perform causal discovery, i.e., establish the existing connections between the variables in each dataset. Through an iterative process of improving the forecasting of a target’s value, we evaluate whether the forecasting improves by adding information from a new potential driver; if so, we state that the driver causally affects the target. On the oil field related datasets, the results we obtained based on causal analysis agree with the interwell connections already confirmed by tracer information; for the cases when the tracer data are available, we used it as our ground truth. This agreement between both estimated and confirmed connections provides us the confidence about the effectiveness of our proposed methodology. To our knowledge this is the first time causal analysis using solely production data is employed to discover interwell connections in an oil field related dataset.

The Stochastic Hydrological Forecast Models for Rainfall Predictions to Aid Climate Change Resilience Within the Tana Alluvial Aquifer Zones Using the Arima Box-Jenkins Algorithms

“…To illustrate this, a meteorologist or climatologist using the ARIMA model to predict drought or rainfall in the future shall be forced to make an assumption that that new rainfall figure expected will have been heavily influenced by past rainfall figure, humidity, temperatures, wind velocity, loud density, et cetera, before deciding whether the amount of rainfall thus qualifies for drought year or a wet one. c) the single column data vector may be manipulated to generate several more columns of independent variables to make linear regressions ML algorithms work with the time-series data (Castro et al, 2023). The process is known as auto-regression.…”