Jinyan Lyu

Traditional Enterprise Resource Planning (ERP) systems excel at quantifying budgetary variances but fail to identify their root causes, leaving financial analysts with time-consuming manual investigation. We present Causal-LLM, a hybrid framework that integrates causal discovery algorithms with Large Language Models (LLMs) to automate root cause diagnosis in budgetary variance analysis. Our approach combines constraint-based causal inference to construct financial causal graphs with LLM-powered contextual reasoning to generate human-interpretable explanations. By leveraging a domain-specific Financial Causal Knowledge Graph, Causal-LLM bridges the gap between statistical correlation and genuine causation. Experimental evaluation on 240 labeled variance cases from a real-world enterprise (24 months of data) demonstrates that our framework achieves 0.87 top-1 accuracy in root cause identification (95% CI: [0.82, 0.91]), outperforming traditional statistical methods (0.68), pure LLM approaches (0.76), and standalone causal methods (0.72). The system generates actionable insights with 0.92 explainability scores (inter-rater agreement ICC=0.84), reducing investigation time from an estimated 2–4 hours (mean: 3.2 hours, based on internal workflow estimates from five senior analysts) to under 10 seconds per case. Results are demonstrated on one manufacturing enterprise; generalization to other industries requires domain-specific ontology adaptation.

Background: Telerehabiliation has been shown to have great potential in expanding access to rehabilitation services, enhancing patients’ quality of life, and improving clinical outcomes. Stationary biking exercise can serve as an effective aerobic component of home-based physical rehabilitation programs. Remote monitoring of biking exercise provides necessary safeguards to ensure exercise adherence and safety in patients' homes. The scalability of the current remote monitoring of biking exercise solutions is impeded by the high cost that limits patient access to these services, especially among older adults with chronic health conditions. Objective: The aim of this project was to design and test two low-cost wireless interfaces for the telemonitoring of home-based biking exercise. Methods: We designed an interactive biking system (iBikE) that comprises a tablet PC and a low-cost bike. Two wireless interfaces to monitor the revolutions per minute (RPM) were built and tested. The first version of the iBikE system uses Bluetooth Low Energy (BLE) to send information from the iBikE to the PC tablet, and the second version uses a Wi-Fi network for communication. Both systems provide patients and their clinical teams the capability to monitor exercise progress in real time using a simple graphical representation. The bike can be used for upper or lower limb rehabilitation. We developed two tablet applications with the same graphical user interfaces between the application and the bike sensors but with different communication protocols (BLE and Wi-Fi). For testing purposes, healthy adults were asked to use an arm bike for three separate subsessions (1 minute each at a slow, medium, and fast pace) with a 1-minute resting gap. While collecting speed values from the iBikE application, we used a tachometer to continuously measure the speed of the bikes during each subsession. Collected data were later used to assess the accuracy of the measured data from the iBikE system. Results: Collected RPM data in each subsession (slow, medium, and fast) from the iBikE and tachometer were further divided into 4 categories, including RPM in every 10-second bin (6 bins), RPM in every 20-second bin (3 bins), RPM in every 30-second bin (2 bins), and RPM in each 1-minute subsession (60 seconds, 1 bin). For each bin, the mean difference (iBikE and tachometer) was then calculated and averaged for all bins in each subsession. We saw a decreasing trend in the mean RPM difference from the 10-second to the 1-minute measurement. For the 10-second measurements during the slow and fast cycling, the mean discrepancy between the wireless interface and tachometer was 0.67 (SD 0.24) and 1.22 (SD 0.67) for the BLE iBike, and 0.66 (SD 0.48) and 0.87 (SD 0.91) for the Wi-Fi iBike system, respectively. For the 1-minute measurements during the slow and fast cycling, the mean discrepancy between the wireless interface and tachometer was 0.32 (SD 0.26) and 0.66 (SD 0.83) for the BLE iBike, and 0.21 (SD 0.21) and 0.47 (SD 0.52) for the Wi-Fi iBike system, respectively. Conclusions: We concluded that a low-cost wireless interface provides the necessary accuracy for the telemonitoring of home-based biking exercise.