Both statistical and neural methods have been proposed in the literature to predict healthcare expenditures. However, less attention has been given to comparing predictions from both these methods as well as ensemble approaches in the healthcare domain. The primary objective of this paper was to evaluate different statistical, neural, and ensemble techniques in their ability to predict patients' weekly average expenditures on certain pain medications. Two statistical models, persistence (baseline) and autoregressive integrated moving average (ARIMA), a multilayer perceptron (MLP) model, a long short-term memory (LSTM) model, and an ensemble model combining predictions of the ARIMA, MLP, and LSTM models were calibrated to predict the expenditures on two different pain medications. In the MLP and LSTM models, we compared the influence of shuffling of training data and dropout of certain nodes in MLPs and nodes and recurrent connections in LSTMs in layers during training. Results revealed that the ensemble model outperformed the persistence, ARIMA, MLP, and LSTM models across both pain medications. In general, not shuffling the training data and adding the dropout helped the MLP models and shuffling the training data and not adding the dropout helped the LSTM models across both medications. We highlight the implications of using statistical, neural, and ensemble methods for time-series forecasting of outcomes in the healthcare domain.
Abstract-We present a system for visualizing memory reference traces, the records of the memory transactions performed by a program at runtime. The visualization consists of a structured layout representing the levels of a cache and a set of data glyphs representing the pieces of data in memory being operated on during application runtime. The data glyphs move in response to events generated by a cache simulator, indicating their changing residency in the various levels of the memory hierarchy. Within the levels, the glyphs arrange themselves into higher-order shapes representing the structure of the cache levels, including the composition of their associative cache sets and eviction ordering. We make careful use of different visual channels, including structure, motion, color, and size, to convey salient events as they occur. Our abstract visualization provides a high-level, global view of memory behavior, while giving insight about important events that may help students or software engineers to better understand their software's performance and behavior. I. INTRODUCTIONThe interactions between modern hardware and software systems are increasingly complex which can result in unexpected interactions and behaviors that seriously affect software performance costing time and money. To address this issue, students and software engineers often spend a significant amount of their time understanding performance and optimizing their software.One common performance analysis technique is to track cache activity within an application. This information is usually provided for very coarse time granularity. At best, cache performance is provided for blocks of code or individual functions. At worst, these results are captured for an entire application's execution. This provides only a global view of performance and limits the ability to intuitively understand performance. An alternative to this coarse granularity is to generate a memory reference trace, which can then be run through a cache simulator to produce a fine-grained approximation of the software's actual cache performance.The biggest challenge when using this approach is sifting through the volume of data produced. Even simple applications can produce millions of references, yet this data contains valuable information that needs to be extracted to better understand program performance. The use of statistical methods or averaging simply produces a coarse understanding of software performance, forgoing the detail available in the trace. Static analysis of memory behavior is also possible [1], but limited only to cases where the program behavior can be deduced at compile time.To address these problems, we propose visualizing the simulated cache and the reference trace, allowing developers
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.