The success of any machine learning system is actually heavily skewed towards the ingredients and not the methods of combining them in a recipe: knowing the problem, asking the right questions, thoroughly understanding the data, and combining results from your test algorithms with expert knowledge are all vital. Finding better features in preprocessing also brings massive gains, as does improving the training set. All of these things take time, trial and error feedback, updates, deep testing, and knowledge transfer. We’ve been building our algorithm for a decade and every day we still see gains from years of fine-tuning. The whitepaper here highlights some of our key takeaways from a decade of doing just this.
Leave your name and contact details and we'll get in touch.