Add interactive schedule_o notebook from Databricks

notebooks/NP04_classification.ipynb

@@ -0,0 +1,909 @@
+{
+ "cells": [
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "application/vnd.databricks.v1+cell": {
+     "cellMetadata": {
+      "byteLimit": 2048000,
+      "rowLimit": 10000
+     },
+     "inputWidgets": {},
+     "nuid": "62d4799f-4935-4a2d-8f0a-5f6383b22cf7",
+     "showTitle": false,
+     "tableResultSettingsMap": {},
+     "title": ""
+    }
+   },
+   "outputs": [],
+   "source": [
+    "df1 = spark.read.table(\"prod_curated.irs.990cn120fields\")\n",
+    "df2 = spark.read.table(\"prod_curated.irs.990standardfields\")"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "application/vnd.databricks.v1+cell": {
+     "cellMetadata": {
+      "byteLimit": 2048000,
+      "rowLimit": 10000
+     },
+     "inputWidgets": {},
+     "nuid": "d5410f3d-7463-43f5-8bfb-528d36e80b42",
+     "showTitle": false,
+     "tableResultSettingsMap": {
+      "0": {
+       "dataGridStateBlob": "{\"version\":1,\"tableState\":{\"columnPinning\":{\"left\":[\"#row_number#\"],\"right\":[]},\"columnSizing\":{\"column\":116},\"columnVisibility\":{}},\"settings\":{\"columns\":{}},\"syncTimestamp\":1758734440525}",
+       "filterBlob": null,
+       "queryPlanFiltersBlob": null,
+       "tableResultIndex": 0
+      }
+     },
+     "title": ""
+    }
+   },
+   "outputs": [],
+   "source": [
+    "from pyspark.sql import SparkSession\n",
+    "import pandas as pd\n",
+    "\n",
+    "# Extract (col, dtype) as dicts\n",
+    "df1_schema = {f.name: f.dataType.simpleString() for f in df1.schema.fields}\n",
+    "df2_schema = {f.name: f.dataType.simpleString() for f in df2.schema.fields}\n",
+    "\n",
+    "# Union of all column names\n",
+    "all_cols = set(df1_schema.keys()).union(df2_schema.keys())\n",
+    "\n",
+    "# Build comparison rows\n",
+    "rows = []\n",
+    "for col in sorted(all_cols):\n",
+    "    in_df1 = col in df1_schema\n",
+    "    in_df2 = col in df2_schema\n",
+    "    \n",
+    "    if in_df1 and in_df2:\n",
+    "        flag = \"both\"\n",
+    "    elif in_df1:\n",
+    "        flag = \"old\"\n",
+    "    else:\n",
+    "        flag = \"new\"\n",
+    "    \n",
+    "    rows.append({\n",
+    "        \"column\": col,\n",
+    "        \"in_df\": flag,\n",
+    "        \"dtype_old\": df1_schema.get(col),\n",
+    "        \"dtype_new\": df2_schema.get(col)\n",
+    "    })\n",
+    "\n",
+    "# Convert to pandas for inspection\n",
+    "comparison_df = pd.DataFrame(rows)\n",
+    "\n",
+    "# If you prefer it as a Spark DataFrame:\n",
+    "spark = SparkSession.builder.getOrCreate()\n",
+    "spark_comparison_df = spark.createDataFrame(comparison_df)\n",
+    "\n",
+    "display(comparison_df)\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "application/vnd.databricks.v1+cell": {
+     "cellMetadata": {
+      "byteLimit": 2048000,
+      "rowLimit": 10000
+     },
+     "inputWidgets": {},
+     "nuid": "2634e810-1046-456f-a00e-34db0ca198a2",
+     "showTitle": false,
+     "tableResultSettingsMap": {},
+     "title": ""
+    }
+   },
+   "outputs": [],
+   "source": [
+    "from pyspark.sql import functions as F\n",
+    "from pyspark.sql.window import Window\n",
+    "\n",
+    "from pyspark.ml.feature import VectorAssembler, StandardScaler\n",
+    "from pyspark.ml.clustering import KMeans\n",
+    "\n",
+    "import plotly.express as px"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "application/vnd.databricks.v1+cell": {
+     "cellMetadata": {
+      "byteLimit": 2048000,
+      "rowLimit": 10000
+     },
+     "inputWidgets": {},
+     "nuid": "cc50ff8a-e01c-417d-b926-fecac95265d0",
+     "showTitle": false,
+     "tableResultSettingsMap": {},
+     "title": ""
+    }
+   },
+   "outputs": [],
+   "source": [
+    "grants_per_state_990 = spark.read.table('sandbox_edward.nonprofit_mapping.grants_per_state_990_filers')\n",
+    "grants_per_state_990pf = spark.read.table('sandbox_edward.nonprofit_mapping.grants_per_state_990pf_filers')"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "application/vnd.databricks.v1+cell": {
+     "cellMetadata": {
+      "byteLimit": 2048000,
+      "rowLimit": 10000
+     },
+     "inputWidgets": {},
+     "nuid": "79330be0-c72e-4670-b6bd-b95665af55c8",
+     "showTitle": true,
+     "tableResultSettingsMap": {},
+     "title": "check for EINs in both 990 and 990pf"
+    }
+   },
+   "outputs": [],
+   "source": [
+    "dual_filers = (\n",
+    "    grants_per_state_990.select(\n",
+    "        'FILEREIN', \n",
+    "        'TAXYEAR'\n",
+    "    )\n",
+    "    .join(\n",
+    "        grants_per_state_990pf.select('FILEREIN', 'TAXYEAR'), \n",
+    "        on=['FILEREIN', 'TAXYEAR'],\n",
+    "        how='inner'\n",
+    "    )\n",
+    ")\n",
+    "\n",
+    "display(dual_filers)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "application/vnd.databricks.v1+cell": {
+     "cellMetadata": {
+      "byteLimit": 2048000,
+      "rowLimit": 10000
+     },
+     "inputWidgets": {},
+     "nuid": "3cd453d0-0bac-42d7-b9d3-51a30be32e6b",
+     "showTitle": false,
+     "tableResultSettingsMap": {},
+     "title": ""
+    }
+   },
+   "outputs": [],
+   "source": [
+    "display(grants_per_state_990.filter(F.col('FILEREIN')=='85-0462315'))"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "application/vnd.databricks.v1+cell": {
+     "cellMetadata": {
+      "byteLimit": 2048000,
+      "rowLimit": 10000
+     },
+     "inputWidgets": {},
+     "nuid": "7232db59-693a-43d9-826b-9e6c2a271626",
+     "showTitle": false,
+     "tableResultSettingsMap": {},
+     "title": ""
+    }
+   },
+   "outputs": [],
+   "source": [
+    "display(grants_per_state_990pf.filter(F.col('FILEREIN')=='85-0462315'))"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "application/vnd.databricks.v1+cell": {
+     "cellMetadata": {
+      "byteLimit": 2048000,
+      "rowLimit": 10000
+     },
+     "inputWidgets": {},
+     "nuid": "64686496-b51f-4aad-ad34-f88e2b69cf61",
+     "showTitle": true,
+     "tableResultSettingsMap": {},
+     "title": "drop dual filers"
+    }
+   },
+   "outputs": [],
+   "source": [
+    "grants_per_state_990 = grants_per_state_990.join(\n",
+    "    dual_filers.select(F.col('FILEREIN'), F.col('TAXYEAR')),\n",
+    "    on=['FILEREIN', 'TAXYEAR'],\n",
+    "    how='left_anti'\n",
+    ")\n",
+    "\n",
+    "grants_per_state_990pf = grants_per_state_990pf.join(\n",
+    "    dual_filers.select(F.col('FILEREIN'), F.col('TAXYEAR')),\n",
+    "    on=['FILEREIN', 'TAXYEAR'],\n",
+    "    how='left_anti'\n",
+    ")"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "application/vnd.databricks.v1+cell": {
+     "cellMetadata": {
+      "byteLimit": 2048000,
+      "rowLimit": 10000
+     },
+     "inputWidgets": {},
+     "nuid": "f0f23c24-a091-49d6-bc7b-64596c89ed0a",
+     "showTitle": true,
+     "tableResultSettingsMap": {},
+     "title": "combine 990 & 990pf orgs into one df"
+    }
+   },
+   "outputs": [],
+   "source": [
+    "grants_per_state = grants_per_state_990.union(grants_per_state_990pf).orderBy('FILEREIN', 'TAXYEAR')"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "application/vnd.databricks.v1+cell": {
+     "cellMetadata": {
+      "byteLimit": 2048000,
+      "rowLimit": 10000
+     },
+     "inputWidgets": {},
+     "nuid": "ac11e27e-18a2-41e0-b8a2-2f6889990a02",
+     "showTitle": false,
+     "tableResultSettingsMap": {},
+     "title": ""
+    }
+   },
+   "outputs": [],
+   "source": [
+    "display(grants_per_state)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {
+    "application/vnd.databricks.v1+cell": {
+     "cellMetadata": {},
+     "inputWidgets": {},
+     "nuid": "6b8aee01-6623-4556-a717-9f58d8af4b6e",
+     "showTitle": false,
+     "tableResultSettingsMap": {},
+     "title": ""
+    }
+   },
+   "source": [
+    "##KMeans Clustering"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "application/vnd.databricks.v1+cell": {
+     "cellMetadata": {
+      "byteLimit": 2048000,
+      "rowLimit": 10000
+     },
+     "inputWidgets": {},
+     "nuid": "013375d3-74b9-48a2-9db0-fe70b09c47f5",
+     "showTitle": true,
+     "tableResultSettingsMap": {},
+     "title": "feature engineering"
+    }
+   },
+   "outputs": [],
+   "source": [
+    "# Normalize/scale features\n",
+    "feature_cols = [\"foreign_percentage\", \"max_recipient_state_percentage\", \"total_recipient_states\"]\n",
+    "assembler = VectorAssembler(inputCols=feature_cols, outputCol=\"features_unscaled\")\n",
+    "df_features = assembler.transform(grants_per_state)\n",
+    "\n",
+    "scaler = StandardScaler(inputCol=\"features_unscaled\", outputCol=\"features\", withStd=True, withMean=True)\n",
+    "df_scaled = scaler.fit(df_features).transform(df_features)\n",
+    "\n",
+    "# Create a composite score - optional, may not add value\n",
+    "# max_states = grants_per_state.select(F.max('total_recipient_states')).collect()[0][0]\n",
+    "# grants_per_state = grants_per_state.withColumn(\n",
+    "#     \"composite_score\",\n",
+    "#     0.5 * (1 - F.col(\"max_recipient_state_percentage\")/100) + \n",
+    "#     0.3 * (F.col(\"total_recipient_states\")/max_states) +   \n",
+    "#     0.2 * (F.col(\"foreign_percentage\")/100)\n",
+    "# )\n",
+    "# feature_cols = [\"foreign_percentage\", \"max_recipient_state_percentage\", \"total_recipient_states\", \"composite_score\"]\n",
+    "# assembler = VectorAssembler(inputCols=feature_cols, outputCol=\"features_unscaled\")\n",
+    "# df_features = assembler.transform(grants_per_state)\n",
+    "\n",
+    "# scaler = StandardScaler(inputCol=\"features_unscaled\", outputCol=\"features\", withStd=True, withMean=True)\n",
+    "# df_scaled = scaler.fit(df_features).transform(df_features)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "application/vnd.databricks.v1+cell": {
+     "cellMetadata": {
+      "byteLimit": 2048000,
+      "rowLimit": 10000
+     },
+     "inputWidgets": {},
+     "nuid": "9bedd689-bb2c-4beb-9f7b-1756ba0c99c5",
+     "showTitle": true,
+     "tableResultSettingsMap": {},
+     "title": "clustering"
+    }
+   },
+   "outputs": [],
+   "source": [
+    "# Clustering on all the scaled features\n",
+    "kmeans = KMeans(featuresCol=\"features\", predictionCol=\"cluster\", k=3, seed=42)\n",
+    "model = kmeans.fit(df_scaled)\n",
+    "\n",
+    "# Assign clusters\n",
+    "df_clustered = model.transform(df_scaled)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "application/vnd.databricks.v1+cell": {
+     "cellMetadata": {
+      "byteLimit": 2048000,
+      "rowLimit": 10000
+     },
+     "inputWidgets": {},
+     "nuid": "4f47c4f6-2143-41a7-b921-55cc3405be3a",
+     "showTitle": false,
+     "tableResultSettingsMap": {},
+     "title": ""
+    }
+   },
+   "outputs": [],
+   "source": [
+    "display(df_clustered)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "application/vnd.databricks.v1+cell": {
+     "cellMetadata": {
+      "byteLimit": 2048000,
+      "rowLimit": 10000
+     },
+     "inputWidgets": {},
+     "nuid": "5cce9fe9-0c60-4c5d-971e-1460e813a0fc",
+     "showTitle": false,
+     "tableResultSettingsMap": {},
+     "title": ""
+    }
+   },
+   "outputs": [],
+   "source": [
+    "# df_clustered.write.mode('overwrite').saveAsTable('sandbox_edward.nonprofit_mapping.funding_orgs_local_vs_national_kmeans_with_composite_feature')\n",
+    "df_clustered.write.mode('overwrite').saveAsTable('sandbox_edward.nonprofit_mapping.funding_orgs_local_vs_national_kmeans_without_composite_feature')"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {
+    "application/vnd.databricks.v1+cell": {
+     "cellMetadata": {},
+     "inputWidgets": {},
+     "nuid": "86298a0c-3526-4749-84d8-33c4119da0d8",
+     "showTitle": false,
+     "tableResultSettingsMap": {},
+     "title": ""
+    }
+   },
+   "source": [
+    "##Cluster Summary - With Composite Feature"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "application/vnd.databricks.v1+cell": {
+     "cellMetadata": {
+      "byteLimit": 2048000,
+      "rowLimit": 10000
+     },
+     "inputWidgets": {},
+     "nuid": "702ac066-0d69-47e9-9411-f080d3a541ea",
+     "showTitle": false,
+     "tableResultSettingsMap": {},
+     "title": ""
+    }
+   },
+   "outputs": [],
+   "source": [
+    "df_clustered = spark.read.table('sandbox_edward.nonprofit_mapping.funding_orgs_local_vs_national_kmeans_with_composite_feature')"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {
+    "application/vnd.databricks.v1+cell": {
+     "cellMetadata": {},
+     "inputWidgets": {},
+     "nuid": "3926601d-c18f-4fed-a87e-06b762d61c6e",
+     "showTitle": false,
+     "tableResultSettingsMap": {},
+     "title": ""
+    }
+   },
+   "source": [
+    "Cluster 0 = local/regional<br>\n",
+    "Cluster 1 = international<br>\n",
+    "Cluster 2 = national"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "application/vnd.databricks.v1+cell": {
+     "cellMetadata": {
+      "byteLimit": 2048000,
+      "rowLimit": 10000
+     },
+     "inputWidgets": {},
+     "nuid": "5ae9099f-f596-4565-954a-40035f6eb880",
+     "showTitle": true,
+     "tableResultSettingsMap": {},
+     "title": "summarize clusters by original features"
+    }
+   },
+   "outputs": [],
+   "source": [
+    "summary = (\n",
+    "    df_clustered\n",
+    "    .groupBy(\"cluster\")\n",
+    "    .agg(\n",
+    "        F.count(\"*\").alias(\"count\"),\n",
+    "        F.avg(\"foreign_percentage\").alias(\"avg_foreign_percentage\"),\n",
+    "        F.median(\"foreign_percentage\").alias(\"median_foreign_percentage\"),\n",
+    "        F.min(\"foreign_percentage\").alias(\"min_foreign_percentage\"),\n",
+    "        F.max(\"foreign_percentage\").alias(\"max_foreign_percentage\"),\n",
+    "        F.avg(\"max_recipient_state_percentage\").alias(\"avg_max_state_pct\"),\n",
+    "        F.median(\"max_recipient_state_percentage\").alias(\"median_max_state_pct\"),\n",
+    "        F.min(\"max_recipient_state_percentage\").alias(\"min_max_state_pct\"),\n",
+    "        F.max(\"max_recipient_state_percentage\").alias(\"max_max_state_pct\"),\n",
+    "        F.avg(\"total_recipient_states\").alias(\"avg_distinct_states\"),\n",
+    "        F.median(\"total_recipient_states\").alias(\"median_distinct_states\"),\n",
+    "        F.min(\"total_recipient_states\").alias(\"min_distinct_states\"),\n",
+    "        F.max(\"total_recipient_states\").alias(\"max_distinct_states\"),\n",
+    "    )\n",
+    "    .orderBy(\"cluster\")\n",
+    ")\n",
+    "\n",
+    "display(summary)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "application/vnd.databricks.v1+cell": {
+     "cellMetadata": {
+      "byteLimit": 2048000,
+      "rowLimit": 10000
+     },
+     "inputWidgets": {},
+     "nuid": "def2f16f-982e-448e-9aa8-782aa01c2193",
+     "showTitle": true,
+     "tableResultSettingsMap": {},
+     "title": "create distribution plots for each cluster (feature: foreign percentage)"
+    }
+   },
+   "outputs": [],
+   "source": [
+    "pdf_clustered = df_clustered.toPandas()\n",
+    "\n",
+    "fig_foreign = px.box(\n",
+    "    pdf_clustered,\n",
+    "    x=\"cluster\",\n",
+    "    y=\"foreign_percentage\",\n",
+    "    title=\"Foreign Percentage by Cluster\",\n",
+    "    labels={\"foreign_percentage\": \"Foreign Percentage\", \"cluster\": \"Cluster\"}\n",
+    ")\n",
+    "fig_foreign.show()"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "application/vnd.databricks.v1+cell": {
+     "cellMetadata": {
+      "byteLimit": 2048000,
+      "rowLimit": 10000
+     },
+     "inputWidgets": {},
+     "nuid": "01020c54-f610-42e2-b43c-f2643f98a576",
+     "showTitle": true,
+     "tableResultSettingsMap": {},
+     "title": "create distribution plots for each cluster (feature: max recipient state percentage)"
+    }
+   },
+   "outputs": [],
+   "source": [
+    "fig_max_recipient = px.box(\n",
+    "    pdf_clustered,\n",
+    "    x=\"cluster\",\n",
+    "    y=\"max_recipient_state_percentage\",\n",
+    "    title=\"Max Recipient State Percentage by Cluster\",\n",
+    "    labels={\"max_recipient_state_percentage\": \"Max Recipient State Percentage\", \"cluster\": \"Cluster\"}\n",
+    ")\n",
+    "fig_max_recipient.show()"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "application/vnd.databricks.v1+cell": {
+     "cellMetadata": {
+      "byteLimit": 2048000,
+      "rowLimit": 10000
+     },
+     "inputWidgets": {},
+     "nuid": "170d71cf-cb3f-44d5-bf90-cc9746a3c1d3",
+     "showTitle": true,
+     "tableResultSettingsMap": {},
+     "title": "create distribution plots for each cluster (feature: number of states)"
+    }
+   },
+   "outputs": [],
+   "source": [
+    "fig_total_states = px.box(\n",
+    "    pdf_clustered,\n",
+    "    x=\"cluster\",\n",
+    "    y=\"total_recipient_states\",\n",
+    "    title=\"Total Recipient States by Cluster\",\n",
+    "    labels={\"total_recipient_states\": \"Total Recipient States\", \"cluster\": \"Cluster\"}\n",
+    ")\n",
+    "fig_total_states.show()"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {
+    "application/vnd.databricks.v1+cell": {
+     "cellMetadata": {},
+     "inputWidgets": {},
+     "nuid": "12dc14fa-0066-4fe5-8a99-d9c6d05860aa",
+     "showTitle": false,
+     "tableResultSettingsMap": {},
+     "title": ""
+    }
+   },
+   "source": [
+    "##Cluster Summary - Without Composite Feature"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "application/vnd.databricks.v1+cell": {
+     "cellMetadata": {
+      "byteLimit": 2048000,
+      "rowLimit": 10000
+     },
+     "inputWidgets": {},
+     "nuid": "1545cd83-3e0a-43f9-9719-14d0f12f5dcb",
+     "showTitle": false,
+     "tableResultSettingsMap": {},
+     "title": ""
+    }
+   },
+   "outputs": [],
+   "source": [
+    "df_clustered = spark.read.table('sandbox_edward.nonprofit_mapping.funding_orgs_local_vs_national_kmeans_without_composite_feature')"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "application/vnd.databricks.v1+cell": {
+     "cellMetadata": {
+      "byteLimit": 2048000,
+      "rowLimit": 10000
+     },
+     "inputWidgets": {},
+     "nuid": "f4a8fe26-50d1-4d55-bd18-0313a1d55136",
+     "showTitle": false,
+     "tableResultSettingsMap": {},
+     "title": ""
+    }
+   },
+   "outputs": [],
+   "source": [
+    "display(df_clustered)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "application/vnd.databricks.v1+cell": {
+     "cellMetadata": {
+      "byteLimit": 2048000,
+      "rowLimit": 10000
+     },
+     "inputWidgets": {},
+     "nuid": "6b2a98fa-d5ee-4c8f-8aad-208a83b2d145",
+     "showTitle": false,
+     "tableResultSettingsMap": {},
+     "title": ""
+    }
+   },
+   "outputs": [],
+   "source": [
+    "from pyspark.sql import functions as F\n",
+    "\n",
+    "df_2023 = (\n",
+    "    df_clustered\n",
+    "    .filter(F.col(\"TAXYEAR\") == 2023)\n",
+    "    .withColumn(\"locality\", F.when(F.col(\"cluster\")==0, \"local/regional\").otherwise(F.when(F.col(\"cluster\")==1, \"international\").otherwise(F.when(F.col(\"cluster\")==2, \"national\").otherwise(None))))\n",
+    "    .select(\n",
+    "        \"FILEREIN\",\n",
+    "        \"TAXYEAR\",\n",
+    "        \"FILERUSSTATE\",\n",
+    "        F.col(\"total_grant_value\").alias(\"value_of_grants\"),\n",
+    "        F.col(\"total_grant_count\").alias(\"number_of_grants\"),\n",
+    "        F.col(\"total_recipient_states\").alias(\"number_of_recipient_states\"),\n",
+    "        F.col(\"foreign_percentage\").alias(\"pct_grant_value_foreign\"),\n",
+    "        F.col(\"max_recipient_state_percentage\").alias(\"pct_grant_value_top_state\"),\n",
+    "        F.col(\"top_recipient_state\").alias(\"top_state\"),\n",
+    "        F.col(\"distinct_recipient_states\").alias(\"recipient_states\"),\n",
+    "        \"locality\",\n",
+    "        \"source\",\n",
+    "    )\n",
+    ")\n",
+    "display(df_2023)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "application/vnd.databricks.v1+cell": {
+     "cellMetadata": {
+      "byteLimit": 2048000,
+      "rowLimit": 10000
+     },
+     "inputWidgets": {},
+     "nuid": "2483cf1c-4bed-47f0-91c0-0364e0f0d5da",
+     "showTitle": false,
+     "tableResultSettingsMap": {},
+     "title": ""
+    }
+   },
+   "outputs": [],
+   "source": [
+    "df_2023.write.mode(\"overwrite\").saveAsTable(\"sandbox_edward.nonprofit_mapping.locality_by_granting_activity_segmentation_funding_orgs_taxyear2023\")"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {
+    "application/vnd.databricks.v1+cell": {
+     "cellMetadata": {},
+     "inputWidgets": {},
+     "nuid": "0781ad05-93a0-46fe-9357-48fea2039b81",
+     "showTitle": false,
+     "tableResultSettingsMap": {},
+     "title": ""
+    }
+   },
+   "source": [
+    "Cluster 0 = local/regional<br>\n",
+    "Cluster 1 = international<br>\n",
+    "Cluster 2 = national"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "application/vnd.databricks.v1+cell": {
+     "cellMetadata": {
+      "byteLimit": 2048000,
+      "rowLimit": 10000
+     },
+     "inputWidgets": {},
+     "nuid": "840cdfdb-dbad-4264-b0fc-85bb060ac2aa",
+     "showTitle": true,
+     "tableResultSettingsMap": {},
+     "title": "summarize clusters by original features"
+    }
+   },
+   "outputs": [],
+   "source": [
+    "summary = (\n",
+    "    df_clustered\n",
+    "    .groupBy(\"cluster\")\n",
+    "    .agg(\n",
+    "        F.count(\"*\").alias(\"count\"),\n",
+    "        F.avg(\"foreign_percentage\").alias(\"avg_foreign_percentage\"),\n",
+    "        F.median(\"foreign_percentage\").alias(\"median_foreign_percentage\"),\n",
+    "        F.min(\"foreign_percentage\").alias(\"min_foreign_percentage\"),\n",
+    "        F.max(\"foreign_percentage\").alias(\"max_foreign_percentage\"),\n",
+    "        F.avg(\"max_recipient_state_percentage\").alias(\"avg_max_state_pct\"),\n",
+    "        F.median(\"max_recipient_state_percentage\").alias(\"median_max_state_pct\"),\n",
+    "        F.min(\"max_recipient_state_percentage\").alias(\"min_max_state_pct\"),\n",
+    "        F.max(\"max_recipient_state_percentage\").alias(\"max_max_state_pct\"),\n",
+    "        F.avg(\"total_recipient_states\").alias(\"avg_distinct_states\"),\n",
+    "        F.median(\"total_recipient_states\").alias(\"median_distinct_states\"),\n",
+    "        F.min(\"total_recipient_states\").alias(\"min_distinct_states\"),\n",
+    "        F.max(\"total_recipient_states\").alias(\"max_distinct_states\"),\n",
+    "    )\n",
+    "    .orderBy(\"cluster\")\n",
+    ")\n",
+    "\n",
+    "display(summary)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "application/vnd.databricks.v1+cell": {
+     "cellMetadata": {
+      "byteLimit": 2048000,
+      "rowLimit": 10000
+     },
+     "inputWidgets": {},
+     "nuid": "3e9a4fb8-11e3-4734-8d8e-943e03e3b738",
+     "showTitle": true,
+     "tableResultSettingsMap": {},
+     "title": "create distribution plots for each cluster (feature: foreign percentage)"
+    }
+   },
+   "outputs": [],
+   "source": [
+    "pdf_clustered = df_clustered.toPandas()\n",
+    "\n",
+    "fig_foreign = px.box(\n",
+    "    pdf_clustered,\n",
+    "    x=\"cluster\",\n",
+    "    y=\"foreign_percentage\",\n",
+    "    title=\"Foreign Percentage by Cluster\",\n",
+    "    labels={\"foreign_percentage\": \"Foreign Percentage\", \"cluster\": \"Cluster\"}\n",
+    ")\n",
+    "fig_foreign.show()"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "application/vnd.databricks.v1+cell": {
+     "cellMetadata": {
+      "byteLimit": 2048000,
+      "rowLimit": 10000
+     },
+     "inputWidgets": {},
+     "nuid": "42bc6e79-cc9f-4745-9751-03c9223a3642",
+     "showTitle": true,
+     "tableResultSettingsMap": {},
+     "title": "create distribution plots for each cluster (feature: max recipient state percentage)"
+    }
+   },
+   "outputs": [],
+   "source": [
+    "fig_max_recipient = px.box(\n",
+    "    pdf_clustered,\n",
+    "    x=\"cluster\",\n",
+    "    y=\"max_recipient_state_percentage\",\n",
+    "    title=\"Max Recipient State Percentage by Cluster\",\n",
+    "    labels={\"max_recipient_state_percentage\": \"Max Recipient State Percentage\", \"cluster\": \"Cluster\"}\n",
+    ")\n",
+    "fig_max_recipient.show()"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "application/vnd.databricks.v1+cell": {
+     "cellMetadata": {
+      "byteLimit": 2048000,
+      "rowLimit": 10000
+     },
+     "inputWidgets": {},
+     "nuid": "b05a45f8-4d81-4ac2-9cfb-b73e18d2051f",
+     "showTitle": true,
+     "tableResultSettingsMap": {},
+     "title": "create distribution plots for each cluster (feature: number of states)"
+    }
+   },
+   "outputs": [],
+   "source": [
+    "fig_total_states = px.box(\n",
+    "    pdf_clustered,\n",
+    "    x=\"cluster\",\n",
+    "    y=\"total_recipient_states\",\n",
+    "    title=\"Total Recipient States by Cluster\",\n",
+    "    labels={\"total_recipient_states\": \"Total Recipient States\", \"cluster\": \"Cluster\"}\n",
+    ")\n",
+    "fig_total_states.show()"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "application/vnd.databricks.v1+cell": {
+     "cellMetadata": {},
+     "inputWidgets": {},
+     "nuid": "3cf19964-c147-4cf4-b7b9-1f31a2e6a256",
+     "showTitle": false,
+     "tableResultSettingsMap": {},
+     "title": ""
+    }
+   },
+   "outputs": [],
+   "source": []
+  }
+ ],
+ "metadata": {
+  "application/vnd.databricks.v1+notebook": {
+   "computePreferences": {
+    "hardware": {
+     "accelerator": null,
+     "gpuPoolId": null,
+     "memory": null
+    }
+   },
+   "dashboards": [],
+   "environmentMetadata": {
+    "base_environment": "",
+    "environment_version": "2"
+   },
+   "inputWidgetPreferences": null,
+   "language": "python",
+   "notebookMetadata": {
+    "pythonIndentUnit": 4
+   },
+   "notebookName": "(Clone) NP04_classification",
+   "widgets": {}
+  },
+  "language_info": {
+   "name": "python"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 0
+}