Systems and methods for high dimensional 3D data visualization
Data visualization processes can utilize machine learning algorithms applied to visualization data structures to determine visualization parameters that most effectively provide insight into the data, and to suggest meaningful correlations for further investigation by users. In numerous embodiments, data visualization processes can automatically generate parameters that can be used to display the data in ways that will provide enhanced value. For example, dimensions can be chosen to be associated with specific visualization parameters that are easily digestible based on their importance, e.g. with higher value dimensions placed on more easily understood visualization aspects (color, coordinate, size, etc.). In a variety of embodiments, data visualization processes can automatically describe the graph using natural language by identifying regions of interest in the visualization, and generating text using natural language generation processes. As such, data visualization processes can allow for rapid, effective use of voluminous, high dimensional data sets.
1. A data visualization system, comprising:
at least one processor; and
a memory comprising a data visualization application, where the data visualization application directs the at least one processor to:
obtain data comprising a set of records, where each record has a plurality of data dimensions;
identify a target dimension in the plurality of data dimensions;
generate a set of ranking metrics reflecting impact of non-target dimensions in the plurality of data dimensions to the target dimension;
calculate a set of correlation coefficients reflecting a degree of statistical correlation between each dimension in the plurality of data dimensions;
generate a set of visualization parameters based on the set of ranking metrics and the set of correlation coefficients;
render a visualization of the target dimension and at least one non-target dimension;
identify a plurality of split points on X, Y, and Z axes of the visualization using a shallow decision tree such that the visualization is split into regions; and
generate a description of the visualization describing at least one region using a natural language processing method.
2. The data visualization system of claim 1 , wherein to generate the set of ranking metrics, the data visualization application further directs the at least one processor to utilize a machine learning model.
3. The data visualization system of claim 2 , wherein the machine learning model is a random forest.
4. The data visualization system of claim 3 , wherein the data visualization application further directs the at least one processor to apply a histogram-based splitting prior to applying the random forest.
5. The data visualization system of claim 1 , wherein the data visualization application further directs the at least one processor to:
select a chart type based on whether the target dimension and each of the at least non target dimension is categorical or numerical; and
render the visualization in accordance with the selected chart type.
6. The data visualization system of claim 5 , wherein the chart type is selected from the group consisting of:
scatter plot;
histogram;
swarm plots;
bar chart;
heat map;
topographic map;
geospatial visualization; and
network visualizations.
7. The data visualization system of claim 1 , wherein the correlation coefficients are selected from the group consisting of: Pearson correlations and cosine similarity.
8. A data visualization method, comprising:
obtaining data comprising a set of records, where each record has a plurality of data dimensions;
identifying a target dimension in the plurality of data dimensions;
generating a set of ranking metrics reflecting impact of non-target dimensions in the plurality of data dimensions to the target dimension;
calculating a set of correlation coefficients reflecting a degree of statistical correlation between each dimension in the plurality of data dimensions;
generating a set of visualization parameters based on the set of ranking metrics and the set of correlation coefficients;
rendering a visualization of the target dimension and at least one non-target dimension;
identifying a plurality of split points on X, Y, and Z axes of the visualization using a shallow decision tree such that the visualization is split into regions; and
generating a description of the visualization describing at least one region using a natural language processing method.
9. The data visualization method of claim 8 , wherein generating the set of ranking metrics comprises utilizing a machine learning model.
10. The data visualization method of claim 9 , wherein the machine learning model is a random forest.
11. The data visualization method of claim 10 , further comprising applying a histogram-based splitting prior to applying the random forest.
12. The data visualization method of claim 8 , further comprising:
selecting a chart type based on whether the target dimension and each of the at least non target dimension is categorical or numerical; and
rendering the visualization in accordance with the selected chart type.
13. The data visualization method of claim 12 , wherein the chart type is selected from the group consisting of:
scatter plot;
histogram;
swarm plots;
bar chart;
heat map;
topographic map;
geospatial visualization; and
network visualizations.
14. The data visualization method of claim 8 , wherein the correlation coefficients are selected from the group consisting of: Pearson correlations and cosine similarity.