Navigation: Introduction → Preliminary Work → Creating a Model Using Weka → Results and Implications

Creating a Model Using Weka

Once we built the comma-separated file of all of our hurricane data from Milestone 3, we then used a program called Weka to analyze and compute regressions for the data.

To see a visual example of how we went about inputting our database and generating a model for it using Weka, take a look at the Milestone 4 video:

As shown in the video, this is a summary of the process we went through to generate our models:

• Download and save our data set.
• Open Weka (we used version 3.4), click on the Explorer button.
• Click on Open File... and under File Type choose CSV Data Files. Locate the data set you downloaded and open it.
• Under the Attributes box, check all the variables named ChangeN[var] except the one for which you want to forecast, then click the Remove button. For example, if you wanted to forecast change in intensity, you would remove ChangeNLat, ChangeNLon, ChangeNMaxWind, and ChangeN34Wind.
• Now click the Classify tab. Click the Choose button, then under functions, click LinearRegression--this is just one of many models that Weka can produce, but it is the simplest, quickest, and most accurate for our purposes.
• Click the drop-down box with the variable names in it and choose the variable you want to forecast. For our intensity example, choose (Num) ChangeNInt.
• Click Start. In a few seconds Weka should generate the model and provide correlation and error statistics.

If you do not want to generate the model yourself, we have provided a tool with which you can forecast hurricanes right in your browser. Please visit the Forecast page if you want to use it.

Previous Page - Next Page