By akademiotoelektronik, 15/09/2022

The Amazon Cloud enters the era of the NO Code

AWS teams its artificial intelligence platform of a graphic layer aimed at putting the development of machine learning models in the hands of data analysts.

How can you make any business decision maker directly create and deploy your machine learning models without depending on data scientists and data engineers?This is the problem to which AWS intends to answer by equipping its AI platform, Sagemaker, with a development environment without code.The new service was unveiled by the Amazon Cloud branch on the occasion of Re: Invent 2021, its world event which is held in Las Vegas from November 28 to December 2.Called Sagemaker Canvas, this No Code overlay allows, from one or more data learning sets, to generate a predictive model which can then be re -involved as the data evolution.

Hundreds models

""Sagemaker Canvas has the Sagemaker platform to automatically clean and combine data sources,"" explains Alex Casalboni, Developer Advocate at AWS.""In the background, he creates hundreds of models, compares their results, selects the most efficient, then generates the prediction (s) which result from it.""For connoisseurs, the solution can draw in several types of treatment: binary classification, multiclasses classification, regression or even Time Series Forecasting.""They will allow you to respond to cases of various use: detection of fraud, anticipation of unsubscriptions, stock prediction"", Glad Alex Casalboni.

In e-commerce, Sagemaker Canvas can for example lead to a model allowing to predict the delivery rate according to various criteria: the carrier used, the distance traveled, the day of shipping day...The demonstration of this use case is carried out on the AWS blog.""We used two files.CSV: order history and the catalog of products.Canvas was responsible for merge them, before going to the training stage and then to the predictive analysis "", explains Alex Casalboni, before specifying:"" It is also possible to draw data from Amazon S3 (the servicestorage oriented AWS objects, editor's note) or in a third -party system like snowflake.""

The data preparation required

Obviously, the data ingested by Canvas must be formatted beforehand.In the example mentioned above, file columns.CVS must correspond so that the demonstration works on the criteria used then to achieve the predictive model (namely: carrier, distance, shipping day).Once this data of data is prepared, canvas automates the entire process.

At the end of the race, a graph allows you to go further by visualizing three key indicators of the model's result: the recall, the precision and the harmonic average.Figures which, obviously, will speak above all to Data Scientists (read the article: What KPI to measure the success of an AI project?).

AWS is not the first to position itself in the field of machine learning no code.Among the publishers present in this segment, there are in particular C3.AI and Datarobot which also offer IA platforms from start to finish.As for Microsoft and Google, they also market equivalent approaches on their respective cloud.The first with its Azure Machine Learning Designer development environment, and the second with its automated machine learning service Google Cloud Automl (read the article The automated machine learning will replace the Data Scientist?).

Le cloud d'Amazon entre dans l'ère de l'IA no code

How can you make any business decision maker directly create and deploy your machine learning models without depending on data scientists and data engineers?This is the problem to which AWS intends to answer by equipping its...

I manage my push subscriptions