linking_main_to_subset2.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from splink.duckdb.linker import DuckDBLinker\n",
    "import splink.duckdb.comparison_library as cl\n",
    "from splink.duckdb.blocking_rule_library import block_on\n",
    "from splink.datasets import splink_datasets\n",
    "from utils import format_model_m_and_us,subset_c"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "df = splink_datasets.historical_50k\n",
    "df2 = subset_c"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "settings = {\n",
    "    \"link_type\": \"link_only\",\n",
    "    \"blocking_rules_to_generate_predictions\": [\n",
    "        block_on(\"first_name\"),\n",
    "        block_on(\"surname\"),\n",
    "    ],\n",
    "    \"comparisons\": [\n",
    "        cl.exact_match(\"first_name\"),\n",
    "        cl.exact_match(\"surname\"),\n",
    "        cl.exact_match(\"dob\"),\n",
    "        cl.exact_match(\"birth_place\", term_frequency_adjustments=True),\n",
    "    ],\n",
    "}\n",
    "\n",
    "settings = {\n",
    "    \"link_type\": \"link_only\",\n",
    "    \"blocking_rules_to_generate_predictions\": [\n",
    "        block_on(\"first_name\"),\n",
    "        block_on(\"surname\"),\n",
    "    ],\n",
    "    \"comparisons\": [\n",
    "        cl.levenshtein_at_thresholds(\"first_name\", [2]),\n",
    "        cl.levenshtein_at_thresholds(\"surname\", [2]),\n",
    "        cl.levenshtein_at_thresholds(\"dob\", [2]),\n",
    "        cl.levenshtein_at_thresholds(\"birth_place\", [2]),\n",
    "    ],\n",
    "}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "linker = DuckDBLinker([df,df2], settings)\n",
    "linker.estimate_u_using_random_sampling(max_pairs=1e6)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "blocking_rule_for_training = block_on([\"first_name\", \"surname\"])\n",
    "linker.estimate_parameters_using_expectation_maximisation(blocking_rule_for_training, estimate_without_term_frequencies=True)\n",
    "\n",
    "blocking_rule_for_training = block_on(\"substr(dob, 1, 4)\")  # block on year\n",
    "linker.estimate_parameters_using_expectation_maximisation(blocking_rule_for_training, estimate_without_term_frequencies=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "linker.m_u_parameters_chart()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = linker.save_settings_to_json()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "m_and_us_df = format_model_m_and_us(linker)\n",
    "# m_and_us_df.to_csv('outputs/main_to_subset2.csv')\n",
    "m_and_us_df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.10.13"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}