MadClothing

Inspiration

Women always woke up with a main blowing plight, rapid and numerous decisions come to place in a rush time. Of which, only constitute tiny part of the daily agenda. Crises may then be induced unnoticeably before turning into a real disaster. How would it be the case? According to a scientific finding, humans only granted limited quota to make decisions for whatever its complexity is. Exceeding the limit, human brain is tired to manage with decision making hence easier to make bad decisions as if other muscles in the human body feel painful after intense workout. This phenomenon is called the decision fatigue.

Thus, MadClothing is a customer orientated mobile application dedicated to give the female users dressing advices based on the both users and general wearing behaviors with simple clicks and drags. Our mission is to assist users to minimize trivial decisions that bother them to choose on. Ideally, all women as our users can get rid of bothers and revel in their daily lives.

What it does

MadClothing is a mobile application supported on iOS and Android platform. Basically, our application have the followings to do:

1. Create account and personal data collection
Users need to create an account to login. During registration, the basic information namely weight, height, age and ethnic is collected to build a personal portfolio to help analyze the suitable pair of apparels. Advance setting is also available to refine and personalized the portfolio.

2. Preference test
Upon registration, users need to take the preference test to notify the system their dressing behavior. The test is comprised of 3 sets of 4 dressing combinations for user to rule out dislike ones. Based on the choices, the app deduces the preference according to three aspects specifically color, texture and style. Optional preference on brand name is also available.

3. Clothes statistic
One of the features is that the app take the users current wardrobe, newly bought clothes and purchasing record on MadClothing into calculation of suggestion pairs. On current wardrobe possession, users can put the data on the client database by taking photo of the object, the app automatically finds some similar wearing in the server database, followed by choosing the most matched one. If the results are not accurate, refinement inputs are in place to work out. For the new bought items, using the camera of the phone to scan the barcode of the product gives the data to database if the product is registered.

4. Matching
Entering the occasion and appropriate dress code is the only things that you need to do. The app will check the weather and then automatically generate a wearable outfit with a pair of matching shoes from your online wardrobe. The preview image will be shown up in your phone at once. If you really dislike the matching, a few more options are still available. The app would like to minimize decisions making and matching work for you, while your preference would not be ignored as well. Our analytic mechanism will fine tune your preference and personalize your matching according to your daily choices.

5. Purchasing
Shopping is one of the most engaging activities that most women like to whenever and wherever. Our apps also finds this no resistivity of the pleasure from buying suitable clothes. Thus, we are offering more opportunities for female users to explore on products which may have greater attractiveness to them, meanwhile spending less efforts and time in finding a needle in a haystack.

How we built it

The core of the application is the suggestion algorithm. This is responsible to give an analysis on how to figure out the best or better pairs based on the sorting out criteria namely weather, occasion, purpose and matching criteria specifically the color match, style and texture.

To run the suggestion algorithm, the app first needs to identify the sorting out criteria which is given each object (wearing) in the server database as a property. Meanwhile, constrains on suitable suggestion objects based of the sorting out criteria that are recorded in 2D array specifying different combinations of those criteria hence give availability of the object.

Meanwhile, on the server database stores large amount of wearing and their properties. We summarized the big data (i.e. the general public dressing behaviors) in an adjacency matrix, treating every apparel in the database as a node while using “1” to show correlation between two objects. Similar work is done in the customer portfolio, user preferable pairs are marked in this way according to their preferences and the apparels in hand.

Then, we can construct different combination pairs based on the user’s available wearing. Nodes in the server database is counted by its number of connection to other feasible nodes. The algorithm put all the customer’s combinations to the numbered nodes, calculated the sum of the node on the path.

With combinations with different ratings, the algorithm randomly select one pair from each of four parts base on the 50th, 75th and 87th percentile since this can assure the pair more likely to be adopted is shown to the customer.

When the apps comes to the suggestion part, the suggested item is selected based on the numbered nodes diagram to give the most suitable and widely accepted item according to the big data on public behavior.