At the beginning, we took the collection of company URLs and read the content of each page and extracted the plain text. In addition, we took a screenshot of each website and compressed it to a suitable size.
The plaintext was converted into a vector for better processing.
Both the vector and the compressed screenshot were used as input for our neural networks. All the data was divided into data sets for training, validation and for the actual test. Depending on the content of the plaintext or the image, the output was a branch of industry.
Deeper learning and understanding of AI.
What it does
Prediciting an economic branch out of a company-URL.
How we built it
Based on Python, Anaconda and CoLab.
Challenges we ran into
Getting the plaintext data from websites tooks very long Predicition of our neuronal network needs to be increased
Accomplishments that we're proud of
Catched data from more than 10.000 websites and parsed it successfully Feeding our neuronal network with this data and making predicitons
What we learned
Basic and advanced features of AI programming Basic Python
What's next for Deeptech-AI_Garbage-Collectors
More Hackathons in future