Week 6

19/01/26

We worked on the design for our website using tailwind css and javascript. We wanted to follow a simplistic modern design that is also colourful and vibrant that we planned previously in our functional specification. We started by modifying the forms page and changing its structure to be more compact and user friendly rather than lining up one by one down the page. We also made home buttons that allow the user to navigate between pages and an interactive submit button for the forms page. On top of this we worked on the hope page and getting it to look like our desired mock-up design. We made it a simple circular search bar to enter the url with a submit button that will automatically fill out the user form with the cars details alongside another simple button that allows the user to fill in the form manually with the “QuickQuote” logo being at the front and center of the page. We kept the design clean and simple for ease of use.

Machine Learning

This week we done Some Principal Component analysis in our dataset, we used these results to help us find out impactful features within our system and here is what why found when figuring PCA out:

Here is a diagram on how components combine give what amount of variance there is

Using sklearn.decomposition import PCA we calculated the top 14 principle components

And heres examples of the top 5:

So to put it in terms to understand we composed our interpretation of the all principal components:

• PC1 Vehicle size & value
  • Weight, cylinder capacity, power, vehicle value, length
  • Captures physical and economic characteristics of the car
• PC2 Driver experience & portfolio maturity
  • Age, seniority, years with license, number of policies
  • Represents driver experience and insurance engagement
• PC3 Claims behavior
  • Number of claims, claims history, policies in force
  • Reflects claims frequency and risk behavior
• PC4–PC5 Risk & demographic segmentation
  • Risk type, age, contract duration, second driver
  • Indicates risk profiles
• Higher PCs (PC6–PC14)
  • Area, payment, distribution channel, fuel type

Capture localised, secondary, or noisy effects

The idea of performing PCA was recommended to us by our mentor, He sent us multiple presentations explaining the topic and as one of us were attending the CSC1045: Machine Learning in Context module, we also learned a lot from attending our lectures.