datathonners

Inspiration

why not

What it does

Our project cleans the dataset, extracts certain features to predict the customer propensity to purchase life and health insurance products.

How we built it

We built the project by cleaning the original dataset, handling class imbalance, performing feature selection and finally comparing performances over different model choices and selecting the most suitable one.

Challenges we ran into

Theoretical knowledge may not directly translate to solving practical issues in the real-world context.

The dataset itself was really challenging due to how dirty it was and felt very tedious to clean.

Model selection was not easy due to the variety of algorithms available, as seen from those introduced in the workshop.

Accomplishments that we're proud of

We successfully handled the noisy or incomplete data in the given original dataset. We also managed to adopt effective feature selection techniques which not only reduced dimensionality but also enhanced the efficiency of our model. We also managed to apply different techniques to tackle the extremely imbalanced dataset.

What we learned

We learnt how to apply different machine learning libraries to assist us in the evaluation of dataset and building the model. What we had learnt in class initially was brought to the real workspace through the introduction of a real dataset. Consequently, it let us really learnt how real data is structured and gave us a challenge to try our hand on how to handle it. The exploration of different techniques also allowed us to gain valuable knowledge which will be useful for our future work or career.