Predictive Modelling

Data has previously been considered a competitive advantage. Now it is more of a commodity and is in many cases freely available and widely disseminated. We recognise a problem for organisations due to data overload. By integrating large amounts of data with specialised knowledge of process held by companies and organisations, predictive models can be created in order to target business activities and operations. In other words businesses need to combine their spatial data and knowledge in a way that allows them to manage more effectively.

Beginning with a customised GIS system, Kenex further enhances spatial data to model predictions, using Spatial Pattern Recognition techniques developed by the Canadian Geological Survey. Techniques employed include: weights of evidence, fuzzy logic, and neural networks. Kenex considers that this is where the real value is added to your data, and once converted to digital form and applied to several layers, statistical modelling allows you to predict unknown values from existing data.

For example information sort from predictive modelling could be be:

• Agricultural - locating the most productive land for deer farming
• Horticultural - modelling of ideal climate and soil conditions for grape growth
• Conservation - finding rare occurrences of plants in National Parks
• Environmental - identifying land parcels that may be of significant environmental importance
• Geological - locating the most prospective areas for mineral exploration
• Geotechnical risk - identifying areas prone to landslides or flooding
• Emergency services - areas of particular hazard risk during earthquakes or storms

The development of a predictive model by Kenex includes the following steps:

1. Data compilation: The initial step is about deciding on the data that will be required to run the model and getting this data into a digital form. Today much of the data required for predictive models is already distributed in a digital GIS format and is available either free-of-charge or available through a licence agreement. Kenex organises the digitising of the remaining data and compiles a digital GIS database that will be used for the spatial modelling.

2. Expert knowledge: No model could be completed without expert subject knowledge provided by leading industry experts. Where not available in-house, Kenex contracts these experts to provide critical information about the datasets used for the model, determine the importance of key themes and the intricacies of the subject matter.

3. Data preparation: Themes and datasets are tested using spatial statistics and are converted into the key themes required for the model. This stage is the most time consuming the for modeller and or the industry expert, as it requires scores of different filtering, analytical, statistical, and data preparation techniques.

4. Modelling:  Kenex uses Arc-SDM on an ESRI GIS platform and MI-SDM on a MapInfo GIS platform to run the models and prepare the data themes. As one of the global leaders in predictive modelling we maintain close links with the software developers to ensure that we're all constantly developing the modelling techniques and the science behind them.

5. Final Reporting / Model Output: The modelling process generates a data layer that shows the probability of the event occurring at a given location. This data set can be displayed as a map that high-lights the most predictive areas, it could be applied to a set of land parcels showing which are most likely to have the event modelling occurring in it, or used to rank the importance of existing spatial objects e.g. mineral permits or land use zones. Models are output to the clients specifications. These can range from a complete digital dataset in the form of a GIS through to a large colour coded map showing the prediction values. Kenex also includes a report all our models discussing the results, how they were obtained, and what data was included.