Geographic Information Systems
Deal with geospatial data - data that identifies
the geographic location of features
– Typically coordinates
What is Choropleth mapping?
Choropleth maps are a form of chart which show the values of variables, instead of using bars or columns they use geographic regions on a map.
Display of attribute (non spatial) data associated with spatial entities.
Choropleth maps are charts, they are not analysis.
Special Operations
- Identify spatial related objects
– beside
– enclosed within
– within radius
Buffers
Bands around spatial objects
Insurance Use of GIS
- Existing policies : Show who and where has
policies, what is protected, and for which sum. - Firmographics : This type shows which
businesses are located side by side. e.g. is a
bookshop near a fire station or a firework shop. - Historical loss information : Spatial record of
all losses for the entire observation period. - Location risk index : For customers, each
location features a specific risk index. - Natural disasters : fires, floods, tornadoes,
earthquakes, etc
Artificial Intelligence
AI is behaviour, which if performed by a human
being, would be called intelligent.
– such as the ability to reason
– learn from past experience
– acquire and retain knowledge
AI cannot do general tasks, e.g. common sense. It has increasing success at specialised tasks where the range of possibilities is limited
- Playing games
- Engineering
- Financial analysis
Expert Systems
– performs a task that would otherwise be performed
by a human expert
– reasoning approach
– “storage” of pre-established knowledge
What is an Expert System
- A computer program that emulates the behaviour
of human experts who are solving real problems
associated with a particular domain of knowledge. - Not used by experts, it stores expert knowledge.
– Only in very specific detailed areas
Benefits of Expert Systems
- faster decisions
– foreign exchange dealing - better decision quality
- capture of scarce expertise
– limited number of experts
– expert not in correct place - integration of several opinions
- cheaper control devices
- spread of knowledge
- reliability
Problems with Expert Systems
- Getting knowledge from expert is hard
– Too busy
– Don’t understand the nature of the information
required
– Don’t believe in ES approach - System is frozen in time
– No systematic updating - Expert rules can be inflexible
– IF (salary > 40000) THEN grant loan
Case Based Reasoning
Case based reasoning looks at previous examples. These previous examples are never quite the same so identifying similarity and modifying the solution are critical.
Uses of CBR in Insurance
- Underwriting Decisions:
– If a new policy proposal is like previous examples,
then it can be routinely accepted. - Claims Handling:
– Refer to similar previous examples . - Pricing and Premium Setting:
– Previous examples can be a starting point for
calculations - Fraud Detection
- Regulatory compliance
– If a case is similar to a previously compliant one
Machine Learning
Machine learning is an approach to the field of Al Systems trained to recognize patterns within data to acquire knowledge
Supervised learning
- Supervised learning
– Provide ML algorithm with curated data set - Classification: Predicting to which discrete class
an entity belongs - Regression: Predicting continuous values of an
entity’s characteristic. - Forecasting: Estimation of macro (aggregated)
variables such as total monthly sales - Attribute Importance: Identifying the variables
(attributes) that are the most important for
prediction
Unsupervised learning
- Unsupervised learning uses machine learning
algorithms to analyse and cluster unlabelled data
sets. - Clustering: Finding natural groupings in the
data. - Association models: Analysing “market
baskets” (e.g., novel combinations of the
products that are often bought together in
shopping carts
– Correlation without explanation
e.g. famous beer and nappies example
Advantages of Machine Learning
- Identification of trends and patterns
- Largely automatic, with limited human work
needed
Issues with Machine Learning
- Garbage in/ Garbage out.
– You have the data, but is it quality data?
– Previous processing of data may create bias. - You still need large computing power
– Cost and energy consumption - Results cannot be easily explained
- If circumstances change then your model may be
useless.
