Computer Science
3 years ago

New model selection criteria outperforms others in predicting time series data.

1 view
arXiv (Cornell University)
Minimum Message Length Autoregressive Moving Average Model Order Selection
Zheng Fang, David L. Dowe, Shelton Peiris, Dedi Rosadi

Paper Summary

The study developed a new method, MML87, to choose the best model for time series data. MML87 was compared to other common methods like AIC and BIC. Results showed that MML87 picked models with lower errors, leading to better predictions in different sample sizes.

Related papers

Theme background image
Computer Science
a year ago

New criterion revolutionizes selection of regimes in Hidden Markov Models!

Model-Based Clustering
Bayesian Networks
2 views
Economics, Econometrics and Finance
4 years ago

Bayesian Information Criterion Outperforms AIC in Selecting Price Relationships

Macroeconomic Analysis
Food Prices
Oil Price Shocks
1 view
Economics, Econometrics and Finance
4 years ago

Astrophysical model selection reveals conflicting conclusions, challenging current beliefs.

Financial Volatility
Missing Data
Regularization Methods
1 view