Binning for Other Charts

Of course, this is not just for Boxplot, but also for other chart types as well. For example, here is a Bar chart showing the average monthly income (Y-Axis) for each category of the working years.

Binning for X-Axis

And this is not just for X-Axis, but also for Color and Repeat By (and Y-Axis for Bubble chart).

Here I’ve assigned Monthly Income column to Color, which automatically divides into 5 categories.

With this, I can show the percentage of each Monthly Income groups for each Working Years group. It’s very clear that the majority of employees whose working years are greater than 20 years are making a high income, but looks it doesn’t really matter how many years you work once you pass the 20th years.

We have seen the categorization of numeric columns with ‘Equal Width’ option so far, but there are other options as well.

Binning with ‘Equal Frequency’ Option

Let’s say you get data like this.

This bar chart is showing the sales by country.

The United States is the highest (the most left hand side) and most of the countries’ sales are very small compared to the United States.

Now, we can visualize this data by using Map and assign Sales to Color. This would categorize the Sales’ numeric values into 5 categories by default, and I have set the number of the categories to 10.

The problem is that only the United States (Red color) look different and everything else looks the similar blue color. This makes it harder to see any differences between the countries other than the United States.

We can try using the ‘Equal Frequency’ option for the binning, which will divide the numeric values into 10 groups (in this case) so that each group will have same number of countries. In this data, there are 165 countries, so each group will have about 16 countries.

Binning with ‘Manual’ Option

You can divide the numeric values by setting specific threshold values.

Here, I’ve assigned the ‘Age’ column to X-Axis and showing the percentage of employees by ‘Job Role’ for each Age group.

In the property, I set specific dividing values (30, 40, 50) and added a label text for each group.

Binning with ‘Outlier’ Option

Another cool option for categorizing numeric values is ‘Outlier’.

Here, I’ve used a histogram to visualize the distribution of Mother Age from the US baby data. By using Repeat By, I’m showing the histogram for each Mother Race type (e.g. White, Black, etc.)

Now, if we want to show the outliers, which means that the mothers who gave birth at very early ages or very late ages, I can assign Mother Age column to Color and setup the categorization.

We can see the outliers are in either green (higher side of the outliers) or red (lower side of the outliers).

This ‘Outlier’ categorization is done against the whole data. But if we want to do it for each Mother Race group, we can set ‘Group By’ setting to use ‘Repeat By’.

Now we can see that outlier ranges are different for each Mother Race group.

Bottom N

We can do the opposite, which is to show the bottom 30 cities based on the delay time.

Condition

Also, we can use Condition as the limiting option.

Here, I’m using the ‘Condition’ to show the cities that have departed greater than 5000 flights.

Log

The log function is useful when you are dealing with data whose distribution is skewed towards the small numbers.

Here is a sales data that is visualized by a histogram.

This is a typical sales data distribution with a long tail.

You can select one of the Log functions from the function list and transform the numeric values and see if the data will be normally distributed (or closer to it).

Scatter

These functions are available for other charts as well. For example, I’m using a Scatter chart to show the relationship between Sales and Sales Compensation of the store sales data.

Again, both Sales and Sales Compensation are skewed towards to small numbers, and this makes it harder to see the relationship in the area where most of the data is.

We can apply the log function to both X-Axis and Y-Axis.

And we can see that the two variables (or columns) are highly correlated (0.90) after they are transformed by the log function.

Scatter

And this is supported for other charts and other places like Color and Repeat By as well.

For example, here is a scatter chart showing the relationship between the temperature and the bike rental counts.

By separating this data into Weekday and Weekend with Color, we can see different patterns between the two.

The number of bike rentals in weekends seem to be much more influenced by the temperature than the ones in weekday.

Many Value Columns!

We have gotten this request a lot, our users want to show more than just 5 value columns!

So we have extended the number of the columns to 30!

Grand Total Calculation

This is not a new feature, but we have changed the way to calculate this grand total.

It used to be calculating the grand total numbers before aggregating the data for Pivot, which could be technically correct, but could be different from what users would expect, especially when you want to calculate average, % ratio, etc.

So we changed it so that the grand total calculations are done on top of the data aggregated in the Pivot table.

You can choose which aggregation function to use from the property.

Random Forest with Boruta

Now Boruta is the default. This means it will always build 20 Random Forest models (you can change this setting) and show the variable importance scores as distribution and perform a statistical test to evaluate which variables importance are statistically significant.

You can change this setting back to a single run with a bar chart for the variable importance view from the property.

Linear Regression with Variable Importance

When you are building the Linear Regression model, more often than not, what you really want to know is which variables have more effects on the target variable.

For this, you can go to Importance tab and compare the variables’ effects on a relative scale.

We have made this view available by default, but you can disable it from the property.

This feature builds a lot of models behind the scene to estimate the effects of the variables, so you might want to disable if you are concerned with the performance.

Logistic Regression with Marginal Effect

Trying to understand the coefficients or the odds ratios for your Logistic Regression models is always a bit tricky. This is why we have introduced the Marginal Effect in v5.2.

It turned out that this has benefited many of our users due to its explainability. One value increase in X would increase the probability of getting TRUE for ‘Marginal Effect’ value percentage at average.

So we have made the metric as the default one, but you can change it from the property.

External Regressor

Now you can add ‘External Predictors’ (or Extra Regressors), which can be used as one of the components to forecast the outcome.

For example, let’s say we want to forecast Sales values.

We can simply forecast the Sales based on the past Sales data.

And RMSE is $16,726, which is considered to be as an ‘average’ difference between the forecasted values and the actual values.

Now, we can add another variable (or column) called ‘Marketing’, which I know it has some degree of correlation with Sales, as the External Predictor.

Notice that the difference between the actual values (blue) and the forecasted values (orange) got smaller.

And the RMSE got smaller, too, as you would expect.

We can add multiple External Predictors.

Here, I’m adding a Sales Compensation variable as one of the External Predictors. And the result looks even better than before.

And the RMSE got even smaller, it’s 5,405 now.

Daily Seasonality

We supported Yearly and Weekly seasonalities before, but with this release, we have added Daily seasonality as well.

With this, you can see what hours of the day have more sales, for example.

Additive vs. Multiplicative

There are two types of data. One is additive, which can be considered as the result of adding numbers. This type of data tends to show a linear trend.

Another is multiplicative, which can be considered as the result of the compounding effect with percentage growth. This type of data tends to show an exponential trend.

And if you’ve got a growing business your sales data tends to look more like the latter. And if you want to forecast such data, now you can select ‘Multiplicative’ for the seasonality effect.

Additive:

Here’s how it looks if we build a forecasting model with the ‘Additive’ option for the seasonality effect.

Notice that the difference between the actual values (blue) and the forecasted values (orange) are becoming wider as the time goes by. And this is a good sign that you want to try the ‘Multiplicative’ option.

Multiplicative:

And here’s how it looks if we switch to ‘Multiplicative’ option.

Notice that the difference between the actual values (blue) and the forecasted values (orange) are equally spread across the years.

And this makes the forecasting model quality better. We can compare the evaluation metrics between the additive and the multiplicative options.

Additive:

Multiplicative:

All the metrics for the multiplicative model are better than the ones for the additive model.