Weather data can provide valuable insights on consumption and production patterns for various industries, especially when it comes to consumption and production forecasting based on historic data. With the use of our historic and current weather data, organizations have many different opportunities to train their machine learning and AI tools very quickly, and thus making better predictions, resulting in significant improvements in efficiency and planning as well as in cost savings.

Meteomatics were excited to make its TV Index and other parameters available within its Weather API to Data Scientists working with the “Naturmuseum” (Museum of Nature) St. Gallen, to help the museum better plan its resources by understanding the relationship between weather and visitor numbers.

  • Meteomatics TV Index identifies "bad weather" which might make people watch more TV
  • The index varies between 0 and 1, where 0 describes nice weather conditions and 1, rainy and stormy weather.

The following figure shows an example of how our TV index looks like:

Chart 1: example of our TV Index: the higher the index value, the higher the probability that people stay at home to watch TV.

Meteomatics' TV index provides a correlation between weather and the numbers of people estimated to watch TV. The index includes a combination of weather types (rainfall, visibility, wind speeds, sunshine duration and windchill) that could lead to more people staying in and watching TV as a result of "bad weather".

The Data Scientists working on the project were able to combine Meteomatics TV Index with two years of historic visitor numbers to train the machine learning prediction model, treating weekdays, weekends/holidays differently, using a non -linear regression method to model the relationship (including other weather parameters that also have a role).

The Project Team identified a strong correlation between adverse weather conditions and the number of visitors to the museum (weekends & holidays 0.71 and weekdays & no holidays is weaker at 0.53).

Chart 2: Meteomatics TV Index & Visitor Numbers @ Weekends & Holidays

Chart 3: Meteomatics TV Index & Visitor Numbers @ Weekdays & No Holidays

These insights have transformed the Museum's operational planning, allowing the Naturmuseum St.Gallen to more accurately predict visitor numbers, improve visitors' experience: staffing levels and safety. The forecast allows the museum to improve it’s resource planning and thus, also to save cost. When aligning resources with visitor numbers, not only the museum will benefit, also the visitors’ experience will improve (e.g. enough people working in the cafeteria, sufficient amount of snacks and drinks, etc.).

Chart 4: Number of forecasted visitors versus observed at Naturmuseum St.Gallen

Meteomatics' TV Index is a good (but also simple) example of how Meteomatics supports machine learning methods by correlating weather insights with other secondary data sources to predict demand and footfall. In fact, Meteomatics are seeing companies across industry applying Meteomatics API to their own business demand intelligence, to create or tune insights and improve their own demand forecasts.

Meteomatics specializes in making weather data more useful and relevant to Meteomatics' API customers, by creating weather parameters that more accurately represent use cases for weather data, with indices that look to correlate impacts of weather on human behavior globally, examples include; favorable weather for a BBQ: going to the beach: cycling: skiing, gardening and many others.

In addition to our special parameters, there is of course also the possibility to obtain pure historical weather data (with over 18'000 parameters) for any place in the world and thus create your own indices. Assuming you have the historical consumption data, visitor numbers or repair numbers for a specific location (or for several locations), you can combine them with the corresponding weather data (and possibly with even more data) and train your machine learning in a very fast and efficient way. This will lead to immediate results in terms of better forecasting. Of course, more accurate data leads to better learning and therefore better predictions, which is why Meteomatics data is particularly valuable for efficient and high-quality machine learning. Besides predicting visitor numbers, the opportunities of historic and current weather data for machine learning also include autonomous driving, navigation and flying, retail consumption predictions, AI based risk assessment for insurances, predictive maintenance, energy demand and production forecast, and many more.

Other projects where Meteomatics Data was alredy used to improve maschine learning and AI include energy demand and production forecasting, logistic management (digital fleet management), hospital capacity planning and dynamic risk assessment. If you are interested in learning more about these projects, please contact [email protected].

Visit our Data Shop to obtain historical data for specific parameters, time series and locations directly or contact [email protected] if you are interested in learning more about the potential of historic and current weather data to boost your machine learning quickly and with high quality data. We are happy to advise you to find the perfect data for your project.