Add residential model parity updates

dvc.lock

@@ -5,17 +5,17 @@ stages:
     deps:
     - path: pipeline/00-ingest.R
       hash: md5
-      md5: 29292ee2bef109914c423c9259aa8879
-      size: 22847
+      md5: 816b28ff1c68d17a9082d0dc839a85c0
+      size: 22844
     params:
       params.yaml:
         assessment:
           year: '2024'
           date: '2024-01-01'
-          triad: city
+          triad: north
           group: condo
           data_year: '2023'
-          working_year: '2024'
+          working_year: '2025'
         input:
           min_sale_year: '2015'
           max_sale_year: '2023'
@@ -31,24 +31,24 @@ stages:
     outs:
     - path: input/assessment_data.parquet
       hash: md5
-      md5: b49601e8a812659026c7358d84f5e16b
-      size: 85702121
+      md5: 1acef7f3c22353411bc15a03d7493164
+      size: 85643154
     - path: input/char_data.parquet
       hash: md5
-      md5: d1a30dd51db2985be57548c1498f2533
-      size: 160972976
+      md5: 5be564143ebae5a67e8f44eb93d839dd
+      size: 159013932
     - path: input/condo_strata_data.parquet
       hash: md5
-      md5: 8fe86e0af29431ecb021f101f79789ee
-      size: 40481
+      md5: b5a85462a7f4de94916b228be45ccd75
+      size: 40543
     - path: input/land_nbhd_rate_data.parquet
       hash: md5
       md5: f3ec9627322bd271bf2957b7388aaa34
       size: 3873
     - path: input/training_data.parquet
       hash: md5
-      md5: 9b2510ac885e4fc77928661a012d8821
-      size: 79812730
+      md5: e818848026f6dc6e3d6af9b8d6b34641
+      size: 79923460
   train:
     cmd: Rscript pipeline/01-train.R
     deps:

params.yaml

@@ -168,28 +168,32 @@ model:
       - "prox_num_pin_in_half_mile"
       - "prox_num_bus_stop_in_half_mile"
       - "prox_num_foreclosure_per_1000_pin_past_5_years"
-      - "prox_num_school_in_half_mile"
       - "prox_airport_dnl_total"
       - "prox_nearest_bike_trail_dist_ft"
       - "prox_nearest_cemetery_dist_ft"
       - "prox_nearest_cta_route_dist_ft"
       - "prox_nearest_cta_stop_dist_ft"
       - "prox_nearest_hospital_dist_ft"
       - "prox_lake_michigan_dist_ft"
-      - "prox_nearest_major_road_dist_ft"
       - "prox_nearest_metra_route_dist_ft"
       - "prox_nearest_metra_stop_dist_ft"
       - "prox_nearest_park_dist_ft"
       - "prox_nearest_railroad_dist_ft"
-      - "prox_nearest_secondary_road_dist_ft"
       - "prox_nearest_university_dist_ft"
       - "prox_nearest_vacant_land_dist_ft"
       - "prox_nearest_water_dist_ft"
       - "prox_nearest_golf_course_dist_ft"
+      - "prox_nearest_road_highway_dist_ft"
+      - "prox_nearest_road_arterial_dist_ft"
+      - "prox_nearest_road_collector_dist_ft"
+      - "prox_nearest_road_highway_daily_traffic"
+      - "prox_nearest_road_arterial_daily_traffic"
+      - "prox_nearest_road_collector_daily_traffic"
+      - "prox_nearest_new_construction_dist_ft"
+      - "prox_nearest_stadium_dist_ft"
       - "acs5_percent_age_children"
       - "acs5_percent_age_senior"
       - "acs5_median_age_total"
-      - "acs5_percent_mobility_moved_from_other_state"
       - "acs5_percent_household_family_married"
       - "acs5_percent_household_nonfamily_alone"
       - "acs5_percent_education_high_school"
@@ -203,11 +207,8 @@ model:
       - "acs5_median_household_total_occupied_year_built"
       - "acs5_median_household_renter_occupied_gross_rent"
       - "acs5_percent_household_owner_occupied"
-      - "acs5_percent_household_total_occupied_w_sel_cond"
-      - "acs5_percent_mobility_moved_in_county"
       - "other_tax_bill_rate"
       - "ccao_is_active_exe_homeowner"
-      - "ccao_is_corner_lot"
       - "ccao_n_years_exe_homeowner"
       - "time_sale_year"
       - "time_sale_day"
@@ -217,6 +218,12 @@ model:
       - "time_sale_day_of_month"
       - "time_sale_day_of_week"
       - "time_sale_post_covid"
+      - "shp_parcel_centroid_dist_ft_sd"
+      - "shp_parcel_edge_len_ft_sd"
+      - "shp_parcel_interior_angle_sd"
+      - "shp_parcel_mrr_area_ratio"
+      - "shp_parcel_mrr_side_ratio"
+      - "shp_parcel_num_vertices"
       - "meta_strata_1"
       - "meta_strata_2"
 

reports/_setup.qmd

@@ -1,4 +1,6 @@
 ---
+execute:
+  echo: FALSE
 params:
   run_id: "2024-02-08-dreamy-sam"
   year: "2024"

reports/performance/_model.qmd

@@ -1036,3 +1036,132 @@ model_big_misses_assessment %>%
 ```
 
 :::
+
+## Variance Over Time
+
+These plot shows show trends in the variance of sale price and estimated FMV. Ideally, the model's estimates should have the same variance as the true values (sales) with respect to time.
+
+::: {.panel-tabset}
+
+```{r _model_organize_variance_data}
+training_data_monthly <- training_data_pred %>%
+  filter(!sv_is_outlier) %>%
+  mutate(
+    meta_sale_date = as.Date(meta_sale_date),
+    year = year(meta_sale_date),
+    month = month(meta_sale_date),
+    difference = (pred_card_initial_fmv - meta_sale_price),
+    squared_difference = difference^2
+  ) %>%
+  group_by(year, month) %>%
+  summarize(
+    total_sales = sum(meta_sale_price),
+    total_fmv = sum(pred_card_initial_fmv),
+    variance_sale = var(meta_sale_price),
+    variance_fmv = var(pred_card_initial_fmv),
+    mean_difference = mean(difference),
+    sse = sum(squared_difference),
+    n = n(),
+    .groups = "drop"
+  ) %>%
+  mutate(
+    variance_diff = variance_fmv - variance_sale,
+    date = make_date(year, month),
+    variance_ratio = variance_fmv / variance_sale,
+    percent_sales = n / sum(n) * 100,
+    percent_sse = sse / sum(sse) * 100
+  )
+training_data_monthly_long <- training_data_monthly %>%
+  pivot_longer(
+    cols = c(
+      variance_sale, variance_fmv, percent_sales,
+      percent_sse, variance_diff
+    ),
+    names_to = "Metric",
+    values_to = "Value"
+  )
+```
+
+### Variance Ratio (FMV / Sale Price)
+
+```{r _model_variance_ratio_chart}
+ggplot(training_data_monthly, aes(x = date, y = variance_ratio)) +
+  geom_line() +
+  geom_point() +
+  labs(
+    x = "Date",
+    y = "Variance Ratio"
+  ) +
+  theme_minimal()
+```
+
+### Total FMV and Sale Price Variance
+
+```{r _model_overall_variance_chart}
+ggplot(
+  training_data_monthly_long %>% filter(Metric %in%
+    c("variance_sale", "variance_fmv")),
+  aes(x = date, y = Value, color = Metric)
+) +
+  geom_line() +
+  geom_point() +
+  geom_smooth(method = "loess", se = FALSE) +
+  labs(
+    x = "Month",
+    y = "Variance",
+    color = "Metric"
+  ) +
+  scale_color_discrete(
+    labels = c(
+      "variance_sale" = "Variance of Sale Price",
+      "variance_fmv" = "Variance of FMV"
+    )
+  ) +
+  scale_y_continuous(labels = function(x) {
+    scales::label_scientific()(x) %>%
+      paste0("$", .)
+  }) +
+  theme_minimal()
+```
+
+### Variance Difference (Sale Price - FMV)
+
+```{r _model_variance_diff_chart}
+ggplot(training_data_monthly, aes(x = date, y = variance_sale - variance_fmv)) +
+  geom_line() +
+  geom_point() +
+  geom_smooth(method = "loess", se = FALSE) +
+  labs(
+    x = "Date",
+    y = "Difference in Variance"
+  ) +
+  scale_y_continuous(labels = function(x) {
+    scales::label_scientific()(x) %>%
+      paste0("$", .)
+  }) +
+  theme_minimal()
+```
+
+### Distribution of Sales and SSE
+
+```{r _model_distribution_sales_sse_chart}
+ggplot(training_data_monthly, aes(x = date)) +
+  geom_bar(aes(y = percent_sales, fill = "Sales"),
+    stat = "identity", position = "identity", alpha = 0.5
+  ) +
+  geom_bar(aes(y = percent_sse, fill = "Sum of Square Errors"),
+    stat = "identity", position = "identity", alpha = 0.5
+  ) +
+  scale_fill_manual(
+    values = c("Sales" = "#00BFC4", "Sum of Square Errors" = "#F8766D")
+  ) +
+  labs(
+    x = "Date",
+    y = "Normalized Scale",
+    fill = "",
+  ) +
+  theme_minimal() +
+  theme(legend.position = "bottom")
+```
+
+:::