Hmmm sounds interesting what is it?
If you ever hate typing to test different model on your dataset and having trouble to load your trained model to a web server, you can try Azure ML STUDIO!
Azure Machine Learning Studio is a GUI-based integrated development environment for constructing and operational Machine Learning workflow on Azure. It doesn't require you to type any programming code to train a model because its a drag and drop tool and best of all you can deploy as a web service after experimental.
Remember the previous post? I was going through on machine learning with scikit-learn but with azure ml it will be easier to use.
Let's first build our first machine learning model.
We will be using one of the dataset that are available on azure, Automobile price data.
We will be predicting the price by using one of the well-known algorithm Linear Regression that we have used previously.
Selecting the columns to include in the operations of the experiment.
We will using the relevant fields that will be needed for the operation to predict the price.
Secondly, we will need to clean the data.
This will improves data quality and in doing so, increases overall productivity. Cleaning the data will allow all outdated or incorrect information is gone – leaving you with the highest quality information.
In this scenario, we will be removing the column that have any invalid fields. This is NOT the best practice to clean datasets because we are losing data.
Next splitting of data, test datasets and training datasets.
Afterwards, we will need to edit metadata field for price as labels.
Marking price as the label (also known as the predictable attribute, or target variable) because we want to predict the price.
Next we will be train the model using linear regression. We will select the price column to be the output value in the process.
This is how it should look like.
We then add the score model to to generate predictions using the trained classification or regression model.
Lastly evaluate the model, to evaluate a scored classification or regression model with standard metrics.
The score of the model is 0.82 which is considered good!
That's it you have created your own simple machine learning model!
Now to host your machine learning model on a web server! You can just do it with a click!
and they will automatically convert it into a predictive experiment
After deploying you will be assigned with an api key where you can use your trained model from different platform.
What is more cooler, you own Request Response API Documentation is automatically created for you!
Thats all folks!!! stay tune for more!
Comments