HEARTCOUNT Tutorial for Beginner
This is a guideline for HEARTCOUNT, an analytics that automatically discovers credible, non-obvious, and practicable patterns from data.
A. HEARTCOUNT LogIn
HEARTCOUNT LogIn (suppose you already have an account)1. LogIn Page (Chrome recommended): https://www.heartcount.io/login
2. In case you do not know the password: [Reset Password] -> Enter your email address -> [Send Password Reset Email] -> Check your email
B. A Data Analysis Campaign Creation
You can upload an excel or csv file. The dataset must be in a tidy and tabular format just like the below
Make your first data campaigndownload a sample dataset 1. drag and drop the dataset(excel, csv) into the cloud
2. click the green cloud icon to upload the dataset
3. (optional) change a campaign name
From the 2nd campaign, you can click [create a new campaign] button at the top of the dashboard page
C. KPI Setting
Once data processing is complete, you will see the “data profile/KPI management” page as below.
You can use pretty much every page of HEARTCOUNT without setting a KPI; however, to fully leverage smart pattern discovery feature, you should set a KPI.
KPI Setting1. select a variable to set as a KPI
2-a. for numeric variable, set a KPI unit and save
2-b. for categorical varialbe, set a class(value) to create a ratio KPI for and save
This is what you will see once setting KPIs
Filtering Bar
Every Page has the uniform filtering bar to filter in/out records and/or variables.
Please check this link if you find it difficult to use the Filtering Bar
Below is the menu composition in terms of What, Why, and How questions, which we will go through one by one.
D-a. What: Analytical Dashboard — Drill-Down
You can perform a drill-down analysis against KPIs on Dashboard page.
Dashboard - KPI drill-down1. choose a variable using [select a variable] menu
2. you can choose up-to 2 variables(dimensions)
Dashboard drill-down is for KPIs. However, on a Drill-Down page, you can perform drill-down on any variables using unlimited number of variables(dimensions)D-b. What: Analytical Dashboard — Smart Discovery
There are just too many variable combinations to drill-down a KPI given enough variables. (for 50 variables, the total number of two variable combinations is 1,225)
Using our proprietary SMART DISCOVERY engine, HEARTCOUNT automatically presents the drill-down conditions that could maximize the KPIs.
What: Analytical Dashboard-Smart Discovery1. click the blue [SMART DISCOVERY] arrow in the Dashboard page (it could take a few mins for the smart discovery results to be available).
2. "Drill-Down I" shows one-dimension(variable) result while "Drill-Down II" shows two-dimension result.
3. You can click the drill-down icon at the top-right corner to perform the drill-down using the discovered condition.
E-a. IMAGINE: Visual Discovery — Smart Plot
Smart Plot is one of HEARTCOUNT’s most versatile data visualization features . UI is quite intuitive and follows the standard data visualization grammar.
Smart Plot: How-To
- smart plot detailed how-to
- our latest animation feature
- stories feature for data story-telling
E-b. IMAGINE: Visual Discovery — Other Visualization Features
There are other visualizations features such as Smart Search, Small Multiples, Smart Ranking, Smart Facets.
Once you’re familiar with Smart Plot UI, you shouldn’t have any difficulties in using other Visual Discovery features.
F. Why: Driver Analysis + Explainer
In Driver Analysis, we present all the statistically significant drivers for the KPI. Internally, we perform linear regression for the given KPI (medical cost in the example below) using the original and *derived(_bin, _percentile) variables combination.
by using the derived variables(such as age_bin) which were automatically created, we can find the non-linear patterns as well.
If you find it difficult to interpret the chart and associated statistics (such as R-squared, P-value) then try the “Explainer” page. “Explainer” presents the same results in plain english.
You can access “Explainer” page using either a smart link (image below) or left-side menu.
Explainer How-T1. click the right/left arrow at the top of left-side narrative window to navigate through the patterns.
2. click [Generate New Narrative] to see new explanation for the same pattern.
3. click [Change Visualization] to change the visualization method.- Want to know more about Data Explainer
F. Why: difference analysis
Difference analysis is to compare any two groups and understand statistical differences. For example, you can compare high-performing and low-performing groups to understand where the differences come from.
Difference Analysis
1. select two groups to compare using "+" icon. In the example below, we set "insurance claim=no" as a group A and "insurance claim=yes' as a group B.
2. click [Compare] button
3. result table presents from the most important variables to least important ones in explaining the difference of the two groups.Result Table
• Variable: the variable that explains the difference between two groups. *difference score is shown as well.
• Chart: distribution comparison chart. blue indicates group A and red indicates group B.
• Group A/B Difference: variable characteristics representing each Group. for example, if you see the 3rd row of the table, "bmi: 16.72~26.58" has relative higher ratio of group A than group B.*how the difference score(%) is calculated:
• it is total area((blue + red + purple) -(minus) shared area(purple)
• if there is no shared (purple) area then the difference metrics = 100%
• if shared area occupies 60% of the total area then difference metrics = 40%.
• the less shared area, difference metics becomes bigger.
• the difference metric should be used for understanding relative strength of a variable in explaining the target variable variation.
• check this if you're curious about the underlying algorithmCase Study
• check this out for a case study using difference analysis
G. How: Micro-Segmentation
Micro-Segmentation is for creating the optimization/classification rule-set using either numeric variable for categorical variable. HEARTCOUNT uses decision tree algorithm for Micro-Segmentation.
Optimization Rule for Numeric Variable
1. choose the variable to optimize
2. by default, top 20% and bottom 20% records are automatically selected to classify between high-performing(top 20%) and low-performing(bottom 20%) group
Result Table & Chart
• the table shows the optimization(classification) rule-sets for each group (top 20% and bottom 20%) along with the associated visualization.
• click "i" icon to understand what each metric such as class, purity means. (we recommend you to google "decision tree algorithm" if you want to understand how the algorithm works and basic concept)Case Study
• check this out for a churn prediction case study using micro segmentation
Classification Rule for Categorical Variable1. choose the categorical variable to classify
2. click [RUN] button
if you have question, please contact me at sidney.yang@idk2.co.kr
Other references- HEARTCOUNT youtube site
- HEARTCOUNT blog site
