Inspiration

The smart mobile app designed to connect expat moms in Amsterdam. Amsterdam is changing: locals are leaving, and expats are coming. The city of Amsterdam is becoming more international. Amsterdam region is home to the most expats in the Netherlands: between 80,000 and 110,000 expats!

Motherhood can be challenging and it's not always easy to find women who understand what you are going through, but it's even more complicated when relocating to a new country.

What it does

MAMS is an intelligent mobile app designed to bring expat moms in Amsterdam together. Matches are made through a clustering algorithm. MAMS is a safety net for expat moms to feel welcome, supported and understood.In addition, it is also a tool that helps moms build their community.

Through the app, moms can register and answer questions about themselves and write a short bio on their profile; this way, the algorithm will find and recommend other moms who, e.g. share the same interests and have children of similar ages.

How we built it

We created a CSV file of randomly generated keys to the questions asked in the app for 1000 fictitious users. A dictionary was created to assign numeric keys to the variables.

This file was used in step 1 of the k-means algorithm to determine the number of clusters to be developed by viewing them on an elbow-graph. On the basis of the data, we identified the right number of clusters should be 3.

In step 2, the k-means algorithm was further applied to assign the fictitious users to one of the 3 defined clusters. The file was amended to include a column of the assigned clusters.

Using the initial file, we manually added a column to include the user bios. The bios had been preprocessed to eliminate stop words, punctuation and other special characters. This modified file was renamed and read into the code. The new column, bios, was then vectorized and a new dataframe was created with the vectorized words. The dataframe with the vectorized bios and the original dataframe were joined into a single dataframe.

The final step was to generate an overview of recommended matches. We used a code to randomly select a user in order to see the top 10 recommended matches for that user along with the match level.

Challenges we ran into

We struggled with connecting the code for preprocessing the Bios field. We wanted the results of the cleaning, vectoring and lemmatising and creating a cosine similarity table to be appended to the output of the code written for assigning users to clusters. Alas, due to time limitations this was not possible but would be something that we'd like to develop down the road.

Accomplishments that we're proud of

We are proud to see our idea come to life!

What we learned

More information to come.

What's next for MAMS app

The MAMS app has potential to build out revenue streams through subscriptions, partnerships, and advertising opportunities as well as branching into other niche target groups and locations.

Built With

Share this project:

Updates