ozitrance

defmodule VlHelper do
  require Explorer.DataFrame
  require Explorer.Series
  alias Explorer.Series
  alias VegaLite, as: Vl
  alias Explorer.DataFrame, as: DF

def get_final_vertical_barriers_chart(df, events_df) do
    df = df
      # Marking first barrier touch rows with true (events_df["event"])

ozitrance

defmodule Labeling do
    ....
  def compute_tripple_barriers do
    ....
    
    ############ This part is mostly for Vega Lite charts ############

    # Stacking our vertical barrier timestamps, and first touches for each event side by side (just like in a table/df) using Nx so we can find the first/lowest one out of all
    stacked_tensors = Nx.stack([events_df["vertical_barrier_timestamp"] |> Series.cast({:f, 64}), events_df["tp"] |> Series.cast({:f, 64}) |> Series.fill_missing(:infinity), events_df["sl"] |> Series.cast({:f, 64}) |> Series.fill_missing(:infinity)], axis: 1)
    # Creating a tensor of the smallest timestamp (first) for each row out of those 3 series - first barrier touch for each event

ozitrance

defmodule Labeling do
    ....
  def compute_tripple_barriers do
    ....
    
        # INSIDE Flow.map
    
    
        # Grabbing the range for the entry event until the vertical barrier
        sliced = DF.slice(df, entry_event..vertical_barrier)

ozitrance

defmodule Labeling do
    ....
  def compute_tripple_barriers do
    .....
    # Calculating Touches and Labeling each trade entry
    events_df =
      # Iterating over each trade entry event and its correspending vertical barrier
      Enum.zip(entry_events_indices |> Nx.to_list, vertical_barriers_indices |> Nx.to_list)
        # Using Flow to run them independently in parallel

ozitrance

defmodule Labeling do
    ....
  def compute_tripple_barriers do
    .....
    # Adding Vertical Barriers for each entry using time offset (commented row below using bars offset)
    vertical_barriers_indices = vertical_barriers(entry_events_indices, df["timestamp"], @vertical_barrier_offset)
    # vertical_barriers_indices = vertical_barriers(entry_events_indices, df["timestamp"], 10, "bars")

    side = 1 # Direction of our trade: 1 for long, -1 for short
    tp = 1.0 # Profit taking multiple

ozitrance

  # Functin defintion with our defaults
  def vertical_barriers(entry_events_indices, timestamps, offset, duration_or_bars \\ "druation")
  
  # If we get text we convert to duration in integer and send it with this argument
  def vertical_barriers(entry_events_indices, timestamps, time_offset, "druation") when is_binary(time_offset) do
    time_offset = ConvertDuration.to_milliseconds(time_offset)
    vertical_barriers(entry_events_indices, timestamps, time_offset, "druation")
  end

  # Where we actually calculate the barriers when using duration as offset

ozitrance

defmodule Labeling do

  require Explorer.DataFrame

  alias Explorer.DataFrame, as: DF
  alias Explorer.Series

  @lookback_window 100 # For our daily_return standard deviation ewma calculation
  @vertical_barrier_offset "60 min"

ozitrance

#include <erl_nif.h>

ERL_NIF_TERM searchsorted(ErlNifEnv* env, int argc, const ERL_NIF_TERM argv[]) {

    if (argc != 3) return enif_make_badarg(env);

    // Getting our offset value    
    int offset;
    if (!enif_get_int(env, argv[1], &offset)) {
        return enif_make_badarg(env);

ozitrance

// Including the Erlang NIF header file to populate our file with all the relevant functions
#include <erl_nif.h>
#include <math.h>
// Creating our function starting with the result type (ERL_NIF_TERM), and the arguments: the Erlang enviorment, argument count, and the arguments array
ERL_NIF_TERM symmetric_cumulative_sum_with_reset(ErlNifEnv* env, int argc, const ERL_NIF_TERM argv[]) {

    // Quick check to make sure we got the amount of arguments were expecting
    if (argc != 2) return enif_make_badarg(env);

    // Getting our threshold value    

ozitrance

def save_to_db(list_of_rows_to_insert) do
  # Falttening our list of lists to one long list
  params = List.flatten(list_of_rows_to_insert)
  # Creating our query with all the placeholders ($1, $2, $3, etc...)
  query = "INSERT INTO submissions 
          # (id, subreddit_id, author_fullname, title, ups, downs, created_utc, selftext, thumbnail, url) 
          # (id, subreddit_id, author_fullname, title, ups, downs, created_utc, selftext, thumbnail, url) 
          # (id, subreddit_id, author_fullname, title, ups, downs, created_utc, selftext, thumbnail, url) 
          # (id, subreddit_id, author_fullname, title, ups, downs, created_utc, selftext, thumbnail, url) 
          # (id, subreddit_id, author_fullname, title, ups, downs, created_utc, selftext, thumbnail, url)