This library was developed in 2 weeks as part of a challenge proposed to us during WebValley 2026. While analysing our challenge we realized that, in order to be able to deal with datasets, MLCasters were required to build a way to visualize the data from scratch. However, the lack of useful and fast plotting tools specifically for MLCast datasets with multidimensional coordinates and different units of measurement make this an inefficient process.
By creating MLCastVis, we have implemented a way to make it easier for anyone to visualize the data present in MLCast.
We thought of a few questions that an MLCaster might have:
- Are there any interesting rain events?
- How does my model perform?
These are the tools that we came up with to help them answer these questions:
- Volatility graph: shows us how much the intensity of the rainfall changes over time, with sudden changes usually corresponding to extreme events
- Comparison grid and interactive comparison plot: two ways to compare the ground truth to the output of a model, one lets us visualize the distribution of rainfall, while the other one provides a more qualitative way to evaluate model performance.
Here is an example of what the volatility plot looks like for the radklim_hourly dataset from July to August 2023:
The peak on July 11 corresponds to the following event:
Here is a link to the repository, which includes a notebook that demonstrates the various functions we have implemented, as well as the reasoning behind their implementation.
Please note that this is a very early version of the library, and still under development. We of course welcome contributions, and would love to see it incorporated into the official MLCast community organisation eventually! Thank you!
This library was developed in 2 weeks as part of a challenge proposed to us during WebValley 2026. While analysing our challenge we realized that, in order to be able to deal with datasets, MLCasters were required to build a way to visualize the data from scratch. However, the lack of useful and fast plotting tools specifically for MLCast datasets with multidimensional coordinates and different units of measurement make this an inefficient process.
By creating MLCastVis, we have implemented a way to make it easier for anyone to visualize the data present in MLCast.
We thought of a few questions that an MLCaster might have:
These are the tools that we came up with to help them answer these questions:
Here is an example of what the volatility plot looks like for the
radklim_hourlydataset from July to August 2023:The peak on July 11 corresponds to the following event:
Here is a link to the repository, which includes a notebook that demonstrates the various functions we have implemented, as well as the reasoning behind their implementation.
Please note that this is a very early version of the library, and still under development. We of course welcome contributions, and would love to see it incorporated into the official MLCast community organisation eventually! Thank you!