OrionAI measures your data before it colours the map

Changes since 2.13.0.

#New

• Classes chosen by measurement. When OrionAI colours areas in classes, it now tries several ways of grouping the values (natural breaks, quantile, equal interval, and for very uneven data head/tail breaks and geometric steps) at three to seven classes, scores each one on how well it fits the data and whether every class is used, and keeps the best. Small sets of areas get fewer classes. It still follows any method or number of classes you ask for.
• A colour key written from the real values. Each row of the key now shows the lowest and highest values actually in that class, so a class never claims values no area has, and a class holding a single area says so, for example "6,750 (1 area)".
• OrionAI checks a field before styling it. It can look at a field's range, spread and zeros to tell a raw count from a rate, or spot values that run either side of zero and need a two-sided scale.
• A cartography check before saving. When OrionAI saves a map it checks for colours that would not match the key, raw counts coloured on areas, rainbow scales on ordered values, empty classes and too many classes, and fixes what it can while it has steps left.

#Improved

• The reason, in one line. The result card says which grouping was used and why, for example "Natural breaks, 5 classes: a few areas are far above the rest, so classes follow the gaps".
• Counts become densities. Ask for "the number of clinics in each district" and OrionAI colours the districts by clinics per km² (or per 1,000 people), so big districts don't look important just for being big, and gives the totals in the summary. It colours raw totals only when you ask for totals rather than a rate.
• Maps compared side by side. Ask for maps of two years or two places with the same classes and OrionAI uses quantile classes so a colour means the same rank on each.