Gamedrop Web App

# Don't forget to start pizzeria
# iex --sname gamedrop --cookie gamedropcookie -S mix phx.server

:ok

Load The Data

We can load our gamedrop database using a migration

alias Gamedrop.Migrations.Seeds
Seeds.run()

Make a Prediction

alias Gamedrop.Ml.Predictor, as: ML
alias Scholar.NaiveBayes.Complement
{model, opts} = ML.model_state()

x = ML.to_x(%{budget: 3, game_types: ["Adventure", "Arcade"]}, opts)
Complement.predict(model, x) |> Nx.to_list() |> List.first()

Pick Top N Predictions

Complement.predict_probability(model, x)
|> Nx.to_flat_list()
|> Enum.with_index()
|> Enum.sort(fn {a, x}, {b, y} ->
  if a == b do
    x < y
  else
    a > b
  end
end)
|> Enum.take(5)

Using Gamedrop

Let’s get the top prediction

alias Gamedrop.Ml.Worker
Worker.predict(%{"budget" => 3, "game_types" => ["Adventure", "Arcade"]})

Or the top 5 predictions.

Worker.predict(%{"budget" => 1, "game_types" => ["Adventure"]}, 5)