A targeted community of researchers investigating biodiversity responses to climate change were approached during May and June to share details of their current research habits and preferences. Their input was collected via the BCCVL scientific survey.
Analysis of the survey results provide the BCCVL project team with valuable insight into the requirements of our primary user base. The following snippets relate to algorithms, models and methods.
- The top 5 ranked algorithms amongst respondents :
- Boosted Regression Trees or Generalised Boosting Model
- Generalised Linear Model
- Generalised Additive Model
- Classification Tree Analysis; Artificial Network; Mahalanobios
- 76% of respondents believe it is either Vital or Important to be able to modify default model settings.
- The majority of respondents use “Random” as the method to select background/pseudo-absence points.
- The majority of respondents use the “Average” ensemble technique.
- Over 80% of respondents indicated AUC was the measure of model accuracy they most often use.