Unscramble Finance

0
7
Unscramble Finance
Unscramble Finance

Unscramble Finance

Online finance is a domain that is full of untapped potential and untapped interest, which is what the love story between Rachel and Mathieu is all about

Nicole Maialboli and Mathieu Lacroix were having a romantic dinner together in Paris, France when a chance encounter transformed their lives.

The couple started dating in 2013 after taking part in a performance of Shakespeare’s Love and Romance. They believe in sharing their love with others, to make it, we need a platform that includes the main aspects that define love. We think of finance as the consumption-based economy. We feel that financial institutions today have fragmented our desire for a richer lifestyle, thus neglecting our values that entail the consumption of finance as “unscramble finance”. We write this project as a metaphor that explores a way for finance to incorporate our needs, our values, and our dreams.

Marie Nilsson, a Brazilian programmer that discovered Unscramble Finance at Inspire Workshop 14. One can study financial market dynamics, create conclusions, and connect them with modern-day feelings like daily relationships in a written way.

Paypal is a business bank that utilizes machine learning for financial payments. PayPal’s dating algorithm determines payment options for the best matching result for each recipient in an instant-ledger fashion.

Find a better matching system for Unscramble Finance that uses machine learning.

We are responsible for creating an extramarital dating algorithm that represents personality and gender diversity that is influenced by the online finance space.

We seek to explore the complex relationship between finance and other sustainable living entities.

We will first define a “Master of Memory” amongst our users that represent their memory capacity and personality within the online finance space. They will find a match that is highly rewarding (equation 1) and will be rewarded with a discount for finding people of similar interests as themselves (equation 2). Using personas from their description, we will also strive to find couples who are compatible. We can apply our knowledge of L2 and L1 knowledge to the graph representation and build a ranking system that reflects data points.

The match rates are determined by a neural network by using the project and dataset of the special interest category. We will use the dataset of HGGC (based on location, gender, age, interests, phone number, and interest group) to find a group of highly correlated users.

The algorithm is first combined with an algorithm made with a boolean key to recognize that a relationship exists in the item. In the event of a match, the population of high-ranked users is selected to have a Reddit chat group that reflects the user’s sentiment.

We leverage “records” that solve this difficulty. These records can be processed and merged with machine learning to correlate this record to the item in its data set. We are using data from a group of smart plants (averaged about 3 times a week) to present data in an impactful and explicit way.

The algorithm decides whether to recommend the items of a data pair.

We discuss compatibility.

We consider the user’s past relationship history to build interest and set a possible profile.

Since we are considering the likes and dislikes, and mostly the age between the participants, we will recommend the less graphic user to get engaged with a more visual person.

The algorithm will recommend the participants for each interaction.

LEAVE A REPLY

Please enter your comment!
Please enter your name here