The architecture diagram demonstrates the end to end pipeline from user input to prediction output. The return value will be parsed by Lambda function again and send the prediction result back to the user. To make this application useful, we will take advantage of AWS API Gateway and Lambda function to build an API taking HTTP POST request that contains airline information(a real-life example could be an APP or website that takes Airline number as input), the POST request will trigger the Lambda function to parse the value and send test data to a Sagemaker endpoint that have the model deployed. Now we can start building our demo on AWS Demo Architecture The processed data ready for training is then stored into an S3 bucket in CSV format. As the purpose of this project is to demonstrate how AWS helps with model training and deployment, We will not spend to much time on how we pre-process the data with Bigquerry(Maybe a future topic). Department of Transportation contains 7.21 million flight records in 2018 with 28 columns.ĭue to a large amount of data(7.21million), we used Google Bigquerry for data cleaning, preprocessing and simple feature engineering. In this article, I would like to demo how we can leverage the power of AWS to build a serverless ML application that predicts air flight delay. With the help of numerous AWS functionalities and tools such as Lambda function, S3, Dynamo DB, the entire process of building a working ML application can be at the click of a mouse. Amazon Sagemaker is one of my favorites, as it largely reduces the effort and hesitation of building, training, and deployment of your models. Luckily, there are many different platforms and tools available to help with model deployment.
#Sagemaker spacenet how to#
Most data enthusiasts know how to build and train a model, but how to deploy your model and make it useful in real-life sometimes can be a challenging issue for beginner data scientists.
0 kommentar(er)
