Dating website optimizer

On dating apps, men & women who have a competitive advantage in photos & texting skills will reap the highest ROI from the app.As a result, I’ve broken down the reward system from dating apps down to a formula, assuming we normalize message quality from a 0 to 1 scale: The better photos/good looking you are you have, the less you need to write a quality message.To model this data, I used two approaches:3-Layer Model: I didn’t expect the three layer model to perform very well.Whenever I build any model, my goal is to get a dumb model working first. I used a very basic architecture:model.add(Convolution2D(64, 3, 3, activation=’relu’))model.add(Max Pooling2D(pool_size=(2,2))) model.add(Flatten())model.add(Dense(128, activation=’relu’))model.add(Dropout(0.5))model.add(Dense(2, activation=’softmax’))The resulting accuracy was about 67%.

As a big-foreheaded, 5 foot 9 asian man who doesn’t take many pictures, there’s fierce competition within the San Francisco dating sphere.

Transfer Learning using VGG19: The problem with the 3-Layer model, is that I’m training the c NN on a SUPER small dataset: 3000 images.

The best performing c NN’s train on millions of images.

One problem I noticed, was I swiped left for about 80% of the profiles. It would difficult to extract information from such a high variation of images.

As a result, I had about 8000 in dislikes and 2000 in the likes folder. Because I have such few images for the likes folder, the date-ta miner won’t be well-trained to know what I like. To fix this problem, I found images on google of people I found attractive. To solve this problem, I used a Haars Cascade Classifier Algorithm to extract the faces from images and then saved it.

Leave a Reply