Emcien Patterns Quickstart

Select Use Case

To get started, choose what use case you would like to do with Emcien Patterns.

For the most part, use cases fall into one of two categories:

1. Find Relationships within a set of data

In this use case, we will use Emcien Patterns to examine a set of bank marketing data to determine what attributes lead to someone making a sale. We will look at different relationships and clusters between the data and illustrate how using Emcien Patterns can enable us to easily and quickly find connections within our data.

2. Make Predictions for a set of data

    In this use case, we will use Emcien Patterns to find the unique patterns and connections in a dataset of bank marketing data with customer attribute data. We will then use that data set to find rules and predictors that would allow us to predict whether or not a sale will be made in a given situation when only having the attribute data. This exercise will enable us to see how easy it is to use the connections that Emcien Patterns finds within data to predict future outcomes.

In this Walkthrough, we will demonstrate both use cases with the same data set. When you know the use case you'd like to continue with, scroll to the top of the page and click on 'Data Prep'.

Data Prep

Click the link below to download the sales data set.

sales-data.csv.zip

Before we move on, let’s take a moment to point out some key attributes about the prepared data:

1. Notice that every row is a separate person. We call this our ‘wide’ format, where each row is a different “transaction” of data

2. Also, see how every column across the top is its own distinct category, and none of the categories are blank or repeat

3. Because the file contained numerical data, those columns were banded using our Banding tool, Bandit. For more details on bandit, see our Advanced Walkthrough.

After the data has been prepared and we understand the traits of good data, scroll to the top of the page and click on 'Automated Analysis'.

Automated Analysis

To begin, let’s go to our home page and click on ‘Analyze Data’

Next, select the file we want to analyze and scroll down

Now, if you are doing Use Case 2 and running a prediction for Sale here, type ‘Sale’ into the predictions field. If you are just trying to find relationships within the data without making predictions, leave this field blank and hit ‘Analyze’.

The engine will have animated steps displaying what it’s doing as it tokenizes, connects, and ranks the data. The picture below describes what is happening in each step.

Now that the data has been analyzed, it's time to see the analyzed product. Scroll to the top of the page and click on 'Review Results'.

Review Results

This is the home page of Emcien Patterns. The colorful chart in the middle is a visual representation of how the data is connected, and gives us an idea of what the engine found in the analysis. To explore the connections within the data a little, let’s go ahead and click on the ‘Clusters’ tab at the top of the page to learn more.

The connections page begins to tell us the story of how our data is connected. The blue boxes at the top represent categories, and are there to show you how items in the selected category are connected with items in other categories. In the picture above, we have selected the category “Age” and we can see how many clusters items in that category make with other categories (a cluster is a connection between 2 or more items). Clicking on a ‘Tell Me’ Button will present each row in a more readable format. Let’s scroll down to find out more about this category.

Further down the page we see the individual connections for items in this category. Each line represents a connection, where the item on the left suggests the presence of the item on the right. So for the first row in our example, when the item “[17.0-32.0] Age” is present in a row, it suggests that the item “single marital status” will be present in that row as well, with a conditional probability of 100%.After reading a few more lines to get an understanding of this category, let’s scroll back to the top and select the ‘Sale’ category to learn more.

After clicking the Sale category, we can see the connections that lead to a ‘Yes’ or ‘No’ for the sale. Clicking the arrow inside of a ‘no’’ box will allow us to go to the category detail page to learn more about the different outcomes of ‘Sale’.

The category detail page shows us the different values for ‘Sale’ that were found in the data set. To learn more about the patterns associated with a successful sale, click the ‘yes’ item.

The item detail screen shows us the other items that are most connected to the item we’ve chosen. What this means is: For the selected item (in this case, ‘yes’ for sale), which other data values occur in transactions most often with our selected item. By seeing this, we can start to see the relation of our item to other values, and get a sense for what is held in the data. To understand this view a bit more, click on ‘Explore Graph’, as seen in the bottom right of the picture.

Here is a graphical representation of the previous screen, showing our selected item (‘yes’ for sale) and all of the items it is most strongly connected with. by highlighting individual items we see the connection patterns and can begin to understand the data within our data set.

Now that we know the patterns and relationships within our sales data, we have finished our first use case. Now, if we'd like to use those connections to find the unique predictors for what causes a successful sale, clikc on 'View Predictors' on the right hand side of your screen.

The Predictions screen shows us the individual combinations of items that, when found in a transaction together, yield a high probability of our predicted item being present in that transaction. Think of these as rules- when the items on the left are found in a row, there is the displayed probability that the item on the right will be present as well. To download a list of these rules, click the ‘Download CSV’ button in the top right.