Unleash the power of Amazon ML

PredicSis.ai - providing the means to truly harness the power of Amazon ML and RedShift to get powerful and meaningful insights from your data

It’s an all-too-familiar story: you have the data, a lot of data. And you’re convinced you can improve your business through exploiting that data, be it better conversion rates, reduced churn or simply a clearer understanding of your customers thanks to Predictive Analytics. There’s no question that having your data to hand in RedShift provides tremendous benefits, nor that Amazon ML simplifies the predictive element.

But once you’ve produced a table using Redshift and Amazon ML, you realise that using multiple tables might not be quite so simple. This is where PredicSis.ai really allows you to harness the power of Redshift and Amazon ML to take you to a level of analysis and understanding that will change the way your business operates.

As an example, let’s imagine you want to improve your email campaigns by better targeting your customers; and that you have 4 data sources for this:

  • Customer reference data
  • Website visited pages
  • Emails received
  • Shopping cart history
Table1: Customer reference table

Table1: Customer reference table

Table 2: Website visited pages

Table 2: Website visited pages

You have the data so ideally you would use it to improve the results. As you can inject only one table into Amazon ML, you have to manually “prepare the data” which involves generating features or aggregates from these 4 tables.

Example of aggregates:

  • Number of emails received over the last 10 days
  • Number of visited pages over the last 5 days

The tricky part is which aggregates to build first?Should I consider 10-day, 5-day, or 2-day timeframes? Should I try all of them? Should I compute means, minimums, maximums, standard deviations? Should I add conditions, like 'Number of visited pages, where page duration is above 1'? And how would I know which ones are impacting and improving the results, or not?

This is a task that only a real expert data scientist can do; it is long and requires many back & forth iterations to find meaningful aggregates. That is, unless you use PredicSis.ai, which can help to transform multiple tables into a single table from which to extract meaningful data out of Redshift. And, importantly, it can do this automatically.

You can feed PredicSis.ai with the 4 tables in RedShift and let it produce a unique enriched data set to pass to Amazon ML. We name it 'automatic feature surfacing'. The overall flow is much faster, accessible, and performance-driven: you stop when you know you've achieved the highest possible result.

We've prepared a small Python script which you can use with your own Redshift cluster. A brief video demonstrates the power of combining PredicSis.ai with Amazon ML.

Use PredicSis.ai to directly inject multiple data sources into Amazon ML and get the most out of your data stored in Redshift. Gain access, in minutes, to best-in-class predictions and to a clear understanding of your customer behaviour; results which would otherwise require expert data science skills to discover, and a few weeks...

Intrigued? If you want to get a sense of PredicSis.ai, we made it easy for you through our demo page. Whilst the demo does not have the multi-table capabilities I mentioned in this post, you'll be able to get a sense of the power of automatic ML