SwiftForecast allows you to forecast the next values in a time series: free, without any software download, and without any mathematical knowledge! Simply insert yout time series in the input field, and our software will take care of calculating the next value, including an estimation of the confidence interval.
How does it work?
SwiftForecast makes use of a large (and growing) collection of different algorithms, from simple linear regressions to more complicated one; by performing a cross-validation, the system is able to automatically detect the most accurate one, and return the best prediction. Also, by weighting the forecasting algorithms according to the number of free parameters in the model, it ensures that overfitting is avoided.
What kind of time series can be forecasted?
Any time series can be introduced in our system! Provided, of course, that it is composed of (real) numbers...
Examples include stock time series for investors, sell or income data for companies, or even your weight!
|General function fits, including:
|ARMA and ARIMA
|Special, propietary algorithms:
Multiple algorithms are combined together to achieve a higher precision.
New look & feel!
We hope you like it. Otherwise, feel free to complain at firstname.lastname@example.org :)
Improved computational time!
While it's still quite high (it's impossible to include a lot of different techniques without some waiting time), the time you'll have to wait has been significantly reduced.
Additionally, we are starting to introduce some new, experimental forecast algorithms. If the results show you algorithms like "WPF", do not worry, the forecast should be quite precise!
Some huge changes, both in the website and the forecasting algorithm!
The most important modification you will notice is in the way results are displayed: no more boring numbers, but a complete graph - thanks to the D3 technology.
We have also improved the way ARMA and ARIMA were calculated - there was some small error before, which prevent showing the correct result. The drawback is that the computational cost has increased: you may have to wait a little more to get the result.
Finally, we have added some sample time series for you to play with (DJI and NASDAQ, with values downloaded in real time!). More to come soon!
After some time, we have uploaded a huge improvement in the system engine. New algorithms have been included, like for instance ARMA and ARIMA; and a new ensemble algorithm, which takes the results of different forecast algorithm to create a new "consensus" result. The validation process has also been improved, providing more reliable estimations of the confidence bound. All of this comes with an increase in the computational cost, but... it's worth it!
Plans for the future include an upgraded web interface, with a better representation of the results obtained... It should be on-line in a couple of weeks.
A minor update of the website, including more information about the system. We will shortly release a new version of the engine, so stay tuned!
Initial version of the SwiftForecast Website and Engine!
Still a lot of work to do in front of us, but we have started and decide to provide you with an initial test version. The engine includes a bunch of standard forecasting techniques based on function fitting: although not state-of-the-art, it provides basic techniques for detecting trends and estimate future values.
As this is just the beginning, we are eager to hear from you! Send any comment or request to email@example.com, we promise to (try to) answer to all of you!