---

Fun Fact:

A recent survey by Slicksdeals.net shows that an average person spent $450 a month on impulsive buy.

A survey done by Creditcards.com shows that 68% of Impulse purchases are done in a physical store

The above infographic by Slickdeals highlights the top 5 most common type of impulse purchases.

Study also shows that a planned shopping trip can reduce impulsive shopping by 13%

---

Background

The Shiny App that will be presented is a Women Shoes Recommender that recommend users a list of shoes based on criteria specified by users.

Shiny App:https://michaelnai.shinyapps.io/Shoes_shiny/

Github: https://github.com/michaelnai/ShinyShoes

Data source: Kaggle https://www.kaggle.com/datafiniti/womens-shoes-prices

Problem Statement

Impulse purchase is the consequence of an unplanned shopping trip and it causes unnecessary expenditure.

Objective

To help planning one's shopping trip for shoes by:

• User specify their preferred feature.
• App return image and information (brand,price,etc…) about the shoes that meet her requirement.
• User can now go straight to the target shop to make purchase, without the need to do unnecessary shopping trip in the mall that will potentially trigger impulse purchase.
• Congratulation you just save yourself some bucks!

---

• Many columns were deemed irrelevant to our usage hence only columns needed for the purpose of creating the App are selected.

• Prices are all normalized to USD.

• Measurement of weights are standardized to KGs.

• NAs in Brand column are imputed with “Others”.

• Observations with NA in the needed columns are removed.

• Observations with NaN in the average price column are removed.

---

• The dataset used was really messy to start with and explanation of the features from the source are not clear.

• Reducing thousands of different colors into a more general scope of colors.

• Some of the features are duplicated record, because two different sites can upload the same shoes. Hence to remove the duplicates we use average to aggregate these features.

• The urls are prepared in a list format (One record can have multiple urls), and to overcome this a proper flatten functions were used

• A lot of outdated URLs that cause images fail to render.