In the last post, I have shown how to install the Azure ML workbench. Also, I have shown an example of classification problem with iris data how to load it and run it
I run the code on the local computer. as you can see in the below picture, the physical place is “local”, there is a possibility to load the run in python and spark as well (will show them later). also after running the scripts, a message has been shown on the right side of the windows that show the status of the running the code (see below picture)
Now, I am going to hover my mouse over the completed job on the right side of the screen.
as you can see in the above picture, there is a summarise information on the run such as the algorithm that has been used, accuracy, and so forth. to have a better view on the run, I am going to click on the run to see the actual results.
By clicking on the result, the “Run Properties” has been shown. The first item is the status of the run (number 3), the start time of run the code, duration, the place that the code has been run, the accuracy of the prediction has been shown.
by scrolling down you will see a section name “Metrics”
In this section, a summary of the performance metrics has been shown. The first one is Precision
Precision = TP / (TP+FP)
So what is Precision?
The below tables shows what is true positive (TP), what is True Negative (TN) and so forth.
so precision tries to show the ratio of cases that algorithm predicts as the positive condition. For instance, for predicting a passenger in Titanic ship are going to survived or not, so for precision, we are going to see how much we are good in predicting the survived cases.
Higher precision is much desire as the above chart shows the chart has an increasing trend for data.
what is the recall?
Recall = TP / (TP+FN)
or it going to look at the ratio of the real data for survival. higher recall better result. or smaller number of False Negative better results.
if you scroll down, you will see some other chart as below that shows the true positive and other in a matrix as below
Finally, at the bottom of the page, you will see the log file for our run as below:
I just click on the “control_log” item and in the below window I able to see the detail of it. Moreover, there is a possibility to download it.
However, besides the above features, there are some places that you able to do data cleaning, data visualization coding and so forth. Just click on the left side of the video, on the database icon, you will see we have two folders: Datastore that is the original dataset we have, the next one is Data preparations that means the data that has been created there.
In next post, I am going to show how we can load data into Azure ML workbench and how to do data cleaning there.
https://en.wikipedia.org/wiki/Precision_and_recall
