multiple-label-alternatives.ipynb

multiple-label-alternatives.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "provenance": [],
      "authorship_tag": "ABX9TyNONstWofDjDfzs2VTqmzyX",
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/gist/firmai/514840e25d2491717ff4c26daebac0eb/multiple-label-alternatives.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "Return Horizon (Eg., 14-day return)"
      ],
      "metadata": {
        "id": "3fqyHLXxcnk-"
      }
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "kR2hXUvxcliN"
      },
      "outputs": [],
      "source": [
        "def return_features(df_pricing, windows=[1,7,14], feature=None):\n",
        "  for w in windows:\n",
        "    df_pricing[f\"return_{feature}_{w}\"] = df_pricing[feature].pct_change(w)\n",
        "  return df_pricing"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "window_returns = [1,3,7,14,28,56,112,224,448]\n",
        "df_pricing = return_features(df_pricing, windows=window_returns, feature=\"closeadj\")\n",
        "df_pricing = return_features(df_pricing, windows=window_returns, feature=\"volume\")"
      ],
      "metadata": {
        "id": "VKDMUqmtcw6P"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "Return Lag (E.g., 14-day return lag 4)"
      ],
      "metadata": {
        "id": "leiGv6NWcprR"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "## rolling returns are correlated, create lags\n",
        "feature = \"closeadj\"\n",
        "for return_type in [1,3,7,14,28,56,112]:\n",
        "  for lag in [1,2,3,4]:\n",
        "    df_pricing[f\"return_{feature}_{return_type}_lag_{lag}\"] = df_pricing[f'return_{feature}_{return_type}'].shift(return_type * lag)"
      ],
      "metadata": {
        "id": "lcE9gZsecytv"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "Rolling Return/Standard Deviation (Eg. 14-day_lag_4_rolling_8_mean/14-day_lag_4_rolling_8_std)"
      ],
      "metadata": {
        "id": "33xaeln1dGtB"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "def col_join(df, win):\n",
        "  df.columns = [\"_\".join(x) for x in df.columns.ravel()]\n",
        "  return df.add_suffix(\"_rolling_{}\".format(win))\n",
        "\n",
        "def rolling_features(df_daily, features=[\"Close\"], windows=[7,14], functions=[\"mean\",\"std\"], method=False, ticker=False):\n",
        "\n",
        "  if method == \"Fast\":\n",
        "    rolling_dfs = [df_daily[features].rolling(i)                  # 1. Create rolling window\n",
        "                  .agg(functions, engine='numba').reset_index(drop=True)          # 2. Apply function\n",
        "                  for i in windows]\n",
        "\n",
        "  rolling_dfs = [col_join(df, win) for df, win in zip(rolling_dfs,windows)] # piece of code to create pretty column names\n",
        "\n",
        "  df_daily = pd.concat((df_daily.reset_index(drop=True), *rolling_dfs), axis=1)\n",
        "  return df_daily\n",
        "\n"
      ],
      "metadata": {
        "id": "v0mGrgStdI5j"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "ticker = \"ticker\"\n",
        "windows = [8]\n",
        "columns = [\"return_closeadj_14_lag_4\"]\n",
        "functions=[\"mean\",\"std\"]\n",
        "\n",
        "df_pricing_features = rolling_features(df_pricing, features=columns, windows=windows, functions=functions, method=\"Fast\", ticker=ticker)\n",
        "\n",
        "df_pricing_features[f\"std_adjusted_return_14\"] = df_pricing_features[\"return_closeadj_14_lag_4_rolling_8_mean\"]/df_pricing_features[\"return_closeadj_14_lag_4_rolling_8_std\"].\n"
      ],
      "metadata": {
        "id": "manLsCP1dUQt"
      },
      "execution_count": null,
      "outputs": []
    }
  ]
}