New study finds day-of-the-week effects in stock market returns using multifractal analysis.
problem Exploring calendar anomalies in stock markets, particularly day-of-the-week effects.
method Multifractal Detrended Fluctuation Analysis (MF-DFA) applied to daily returns of market indices.
result Monday returns exhibit more persistent behavior and richer multifractal structures than other days.
Study finds biotech stocks perform better on Wednesdays and Thursdays.
problem Day-of-the-week effect on biotechnology stocks performance.
method Daily returns analysis using GARCH processes and asymmetric GARCH models.
result Biotechnology stocks have higher returns on Wednesdays, Thursdays, and Fridays.
Study examines Bitcoin's volatility and returns using stochastic volatility model.
problem Characterizing Bitcoin as a financial asset and its volatility patterns.
method Asymmetric stochastic volatility model applied to Bitcoin data from 2013-2019.
result Bitcoin shows weak post-holiday effects and no asymmetry effect in returns and volatility.
Paper presents a new time-series segmentation technique for mobile phone user behavior.
problem Current segmentation techniques do not accurately capture individual user behavior over time.
method Behavior-Oriented Time Segmentation (BOTS) technique that considers temporal coverage and number of incidences.
result BOTS technique better captures user behavior at various times of day and week.
Paper proposes a new test to detect spurious seasonality in time series data.
problem Detecting spurious seasonality in time series data.
method Developed a non-parametric test based on ordinal patterns using symbolic dynamics.
result The day-of-the-week effect is partly an artifact of hidden correlation structure.
Automated trading systems on developed and emerging capital markets are studied in this paper. The standard for developed market is automated trading system with 40-days simple moving average. We tested it for the index SIX Industrial for 1000 and 730 trading days of the slovak emerging capital market. The Buy and Hold…
EGPR method forecasts power grid load and generation with high accuracy.
problem Accurate week-long forecasting of load demand and power generation for efficient power grid operation.
method Gaussian process regression with ensembles of preceding weeks' data.
result EGPR method outperforms traditional methods in forecasting weekly load and generation.
Stock market instability increased after 2018 February.
problem Understanding the recent increase in stock market instability.
method Applied a complex systems model to analyze stock price comovement and U parameter. result The U parameter value has decreased significantly since February 2018, indicating a new regime of market behavior. Study finds recurring patterns in cryptocurrency volatility and liquidity.
problem Recurring patterns in volatility and liquidity of major cryptocurrencies.
method Data from two centralized exchanges and a decentralized exchange analyzed for patterns.
result Systematic patterns in volatility and liquidity across different timeframes.
Predicting and improving player retention is crucial to the success of mobile Free-to-Play games. This paper explores the problem of rapid retention prediction in this context. Heuristic modeling approaches are introduced as a way of building simple rules for predicting short-term retention. Compared to common classifi…
Averaging recent model checkpoints speeds up training time.
problem Training large vision or language models is time-consuming.
method Average the weights of the k latest checkpoints.
result Speeds up training by dozens of epochs, saving up to 68 GPU hours.
Paper quantifies dataset shift for credit card fraud detection.
problem Change in purchase behavior over time affects fraud detection accuracy.
method Measures day-to-day dataset shift using classification efficiency and clustering.
result Improves credit card fraud detection by incorporating dataset shift knowledge.
Spotify improves content mix using contextual bandits.
problem Skewed historical data and varying user preferences across contexts.
method Contextual bandits to dynamically learn optimal content type distribution.
result Improved precision and user engagement with under-represented content types.
Forecasting COVID-19 cases in Senegal using machine learning.
problem Predicting the inflection point and ending time of COVID-19 cases in Senegal.
method Visualization and machine learning techniques applied to public data.
result Forecasted the inflection point and possible ending time of COVID-19 cases in Senegal.
The paper models ATM cash withdrawal chaos and forecasts using deep learning.
problem Forecasting ATM cash withdrawals in an Indian bank.
method Chaos modeling of ATM cash withdrawal time series, deep learning methods (ARIMA, RF, SVR, MLP, GMDH, GRNN, LSTM, 1D CNN).
result Deep learning models show similar performance to random forest in forecasting ATM cash withdrawals.
Machine learning improves electricity price forecasting.
problem Predicting electricity prices in various horizons.
method Application of machine learning techniques to EPF models.
result Machine learning models outperform traditional methods.
The behaviors of patients with depression are usually difficult to predict because the patients demonstrate the symptoms of a depressive episode without a warning at unexpected times. The goal of this research is to build algorithms that detect signals of such unusual moments so that doctors can be proactive in approac…
KataGo accelerates Go self-play learning by 50x.
problem Efficiently learning in large state spaces like Go.
method Improved AlphaZero process and architecture.
result 50x reduction in computation time.
Model uses GAMs to forecast hourly electricity load weeks to one year ahead.
problem Accurate mid-term hourly load forecasting for power plant operation and energy management.
method Generalized Additive Models (GAMs) with P-splines and autoregressive post-processing.
result Significantly enhanced forecasting accuracy compared to state-of-the-art methods.
The paper improves electricity price forecasting using future prices.
problem Reliable short to mid-term forecasts of electricity prices despite high variation.
method Combining econometric autoregressive models with future prices for improved forecasting performance.
result The model can outperform other models in the literature and maintain hourly precision.
Model predicts COVID-19 spread with better accuracy than existing methods.
problem Limited daily samples in time for data-driven methods.
method Integrated spatiotemporal model combining epidemic differential equations and RNN.
result Model outperforms existing methods in forecasting cases.
It is usually assumed that stock prices reflect a balance between large numbers of small individual sellers and buyers. However, over the past fifty years mutual funds and other institutional shareholders have assumed an ever increasing part of stock transactions: their assets, as a percentage of GDP, have been multipl…
SilentPhone identifies opportune times to silence phones to reduce interruptions.
problem Inappropriate phone notifications cause interruptions for users and others.
method Data-driven approach using past phone log data to infer unavailability.
result Identifies opportune moments for call interruptions and generates silent mode rules.
Deep learning models predict call center volumes with seasonal patterns.
problem Forecasting call center volumes with complex seasonal behavior.
method Investigated recurrent neural networks (RNNs) including Elman, LSTM, and GRU models.
result Optimal RNN configurations outperform other forecasting techniques.
High-value transactions between Australian banks are settled in the Reserve Bank Information and Transfer System (RITS) administered by the Reserve Bank of Australia. RITS operates on a real-time gross settlement (RTGS) basis and settles payments sourced from the SWIFT, the Austraclear, and the interbank transactions e…
Paper uses machine learning and SIR models to predict COVID-19 cases.
problem Predicting the spread of COVID-19 cases for control measures.
method Machine learning and SIR models (deterministic and stochastic) with numerical approximations.
result Predictions help in finding concrete actions to control the pandemic.
Using the correlation matrix formalism we study the temporal aspects of the Warsaw Stock Market evolution as represented by the WIG20 index. The high frequency (1 min) WIG20 recordings over the time period between January 2001 and October 2005 are used. The entries of the correlation matrix considered here connect diff…
XferNAS reduces neural architecture search time by 33x.
problem Expensive neural architecture search for new tasks.
method Transfer knowledge from previous tasks to reduce search time.
result Reduction of search time from 200 to 6 GPU days.
Improved county-level COVID-19 forecasting model using LSTM and data augmentation.
problem Accurately forecasting county-level COVID-19 cases to optimize medical resources.
method Adapted TDEFSI-LONLY model, utilized LSTM, data augmentation, and inter-county mixing.
result CLEIR-Net model provides better forecasts than TDEFSI-LONLY.
Gaussian process models improve MJO predictions with better uncertainty quantification.
problem Lack of uncertainty quantification in MJO predictions by machine learning models.
method Developed a nonparametric strategy based on Gaussian process models, calibrating them using empirical correlations and proposing a posteriori covariance correction.
result Gaussian process models provide better prediction skills and extended probabilistic coverage for MJO forecasts.
Study of historic stock returns distributions, highlighting asymmetry and outliers.
problem Understanding the asymmetry in accumulated gains and losses in stock returns over time.
method Analyzing decades-long historic distributions of S&P500 returns, comparing gains and losses, using statistical U-tests and fitting log-log scale linearly.
result The mean of de-trended distributions increases linearly with the number of days of accumulation, and the overall skew is negative, indicating heavier tails of losses.
Transformer model forecasts electricity price spread for virtual bidding.
problem Volatility in renewable energy causes price forecasting challenges.
method Transformer-based deep learning model using various time-series features.
result Trading strategy at peak hour yields nearly consistent profit.
HARNet improves volatility forecasting using deep neural networks.
problem Lack of deep learning in volatility forecasting.
method HARNet based on dilated convolutional layers, explicitly initialized to match HAR model.
result HARNet significantly improves forecasting accuracy compared to HAR models.
This work proposes a scalable framework for trusted multi-party computations using blockchain.
problem Ensuring trust in results from multi-agent computational experiments.
method Combining distributed validation and blockchain for immutable audits, reducing storage and communication costs.
result Guaranteed verifiability and validity of local computations in a scalable multi-agent environment.
Favorit strategy helps farmers mitigate market price fluctuations.
problem Mitigating adverse impact of price fluctuation on farmers.
method Analyzes historical price data to select optimal market timing for crops.
result Developed a strategy to reduce volatility risk for Indian farmers.
Bayesian optimization has become a successful tool for hyperparameter optimization of machine learning algorithms, such as support vector machines or deep neural networks. Despite its success, for large datasets, training and validating a single configuration often takes hours, days, or even weeks, which limits the ach…
Study analyzes online student behavior patterns using log data.
problem Understanding and optimizing student learning in online educational systems.
method Non-negative matrix factorization techniques for soft clustering.
result Behavioral changes of individual students and the system over time.
A new psychotherapy dialogue model learns compressed text representations.
problem Modeling psychotherapy dialogues with small datasets.
method Nonparametric kernel functions and hashcode representations.
result Significantly outperforms state-of-the-art models in psychotherapy sessions.
Optimizing data movement significantly improves transformer training efficiency.
problem Data movement is a major bottleneck in training transformers.
method Developed a recipe to globally optimize data movement in transformers.
result Achieved up to 1.30x performance improvement over state-of-the-art frameworks.
The p-index improves investment performance for NYSE stocks but not for SSE stocks.
problem Improving investment performance for stocks using the p-index.
method Comparing different p-ratio strategies and empirical efficient frontiers for SSE and NYSE stocks.
result The p-index enhances investment performance for NYSE stocks but not for SSE stocks.
Efficiently trains BERT on academic GPUs in 12 days.
problem Training large-scale BERT models is expensive and time-consuming.
method Optimizes training on multiple GPUs and nodes, reducing costs.
result Trains BERT on academic GPUs in 12 days, not requiring expensive hardware.
New econometric results for financial duration models under varying tail behaviors.
problem Estimation and inference challenges in financial durations models with random event counts.
method Analysis of likelihood estimators for ACD models, focusing on tail behavior and stationarity.
result Asymptotic normality breaks down for tail indices smaller than one, leading to mixed Gaussian estimators with non-standard rates of convergence.
Paper proves equivalence between time consistency and nested formula in financial rankings.
problem Ranking consistency of stochastic processes over time.
method Minimalist definition of Time Consistency and proof of equivalence with Nested Formula.
result Two assessments are consistent if one is factored into the other.
A Bayesian framework models dynamic probability predictions over time.
problem Dynamic probability predictions over time in various settings.
method Gaussian latent information martingale (GLIM) framework.
result GLIM outperforms baseline methods in predicting future uncertainties.
Investigates statistical properties and multifractality of Bitcoin prices.
problem Analyzing statistical and multifractal properties of Bitcoin prices.
method Examined 1-min returns of Bitcoin prices, used multifractal detrended fluctuation analysis, and applied GARCH models.
result Bitcoin exhibits multifractality due to both temporal correlation and fat-tailed distribution.
Study detects Bitcoin bubbles and predicts crashes using adaptive multilevel time series detection.
problem Detecting and predicting Bitcoin price bubbles and crashes.
method Adaptive multilevel time series detection based on LPPLS model.
result LPPLS confidence indicator provides effective warnings for bubble detection and crash prediction.
Neural networks improve VaR estimation accuracy and robustness.
problem Estimating Value at Risk (VaR) in financial markets.
method Generative regime switching framework with Monte-Carlo simulations, neural networks initialized via best model, balanced incentive function, reduced training data.
result Neural networks outperform traditional methods in VaR estimation, especially with less data.
This study assesses the reproducibility of 1H-MRS scans across different vendors and sessions.
problem Lack of harmonization in magnetic resonance spectroscopy protocols among vendors.
method Analysis of CV and ICC for within- and between-sessions, and correlation coefficients for across machines.
result Metabolite concentrations are highly reproducible across different vendors and sessions.