Regression¶
This tutorial uses safeds on house sales data to predict house prices.
- Load your data into a
Table, the data is available underdocs/tutorials/data/pricing.csv:
In [1]:
Copied!
from safeds.data.tabular.containers import Table
pricing = Table.from_csv_file("data/house_sales.csv")
# For visualisation purposes we only print out the first 15 rows.
pricing.slice_rows(0,15)
from safeds.data.tabular.containers import Table
pricing = Table.from_csv_file("data/house_sales.csv")
# For visualisation purposes we only print out the first 15 rows.
pricing.slice_rows(0,15)
Out[1]:
| id | year | month | day | zipcode | latitude | longitude | sqft_lot | sqft_living | sqft_above | sqft_basement | floors | bedrooms | bathrooms | waterfront | view | condition | grade | year_built | year_renovated | sqft_lot_15nn | sqft_living_15nn | price | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 0 | 0 | 2014 | 5 | 2 | 98001 | 47.3406 | -122.269 | 9397 | 2200 | 2200 | 0 | 2.0 | 4 | 2.50 | 0 | 1 | 3 | 8 | 1987 | 0 | 9176 | 2310 | 285000 |
| 1 | 1 | 2014 | 5 | 2 | 98003 | 47.3537 | -122.303 | 10834 | 2090 | 1360 | 730 | 1.0 | 3 | 2.50 | 0 | 1 | 4 | 8 | 1987 | 0 | 8595 | 1750 | 285000 |
| 2 | 2 | 2014 | 5 | 2 | 98006 | 47.5443 | -122.177 | 8119 | 2160 | 1080 | 1080 | 1.0 | 4 | 2.25 | 0 | 1 | 3 | 8 | 1966 | 0 | 9000 | 1850 | 440000 |
| 3 | 3 | 2014 | 5 | 2 | 98006 | 47.5746 | -122.135 | 8800 | 1450 | 1450 | 0 | 1.0 | 4 | 1.00 | 0 | 1 | 4 | 7 | 1954 | 0 | 8942 | 1260 | 435000 |
| 4 | 4 | 2014 | 5 | 2 | 98006 | 47.5725 | -122.133 | 10000 | 1920 | 1070 | 850 | 1.0 | 4 | 1.50 | 0 | 1 | 4 | 7 | 1954 | 0 | 10836 | 1450 | 430000 |
| 5 | 5 | 2014 | 5 | 2 | 98007 | 47.6022 | -122.134 | 6700 | 1570 | 1570 | 0 | 1.0 | 3 | 1.50 | 0 | 1 | 4 | 7 | 1956 | 0 | 7300 | 1570 | 419000 |
| 6 | 6 | 2014 | 5 | 2 | 98008 | 47.6188 | -122.114 | 8030 | 2000 | 1000 | 1000 | 1.0 | 3 | 2.25 | 0 | 1 | 4 | 8 | 1963 | 0 | 8250 | 2070 | 420000 |
| 7 | 7 | 2014 | 5 | 2 | 98011 | 47.7698 | -122.222 | 9655 | 2210 | 1460 | 750 | 1.0 | 5 | 2.50 | 0 | 1 | 3 | 8 | 1976 | 0 | 8633 | 2080 | 470000 |
| 8 | 8 | 2014 | 5 | 2 | 98011 | 47.7419 | -122.205 | 12261 | 2730 | 2730 | 0 | 2.0 | 4 | 2.50 | 0 | 1 | 3 | 9 | 1991 | 0 | 10872 | 2730 | 612500 |
| 9 | 9 | 2014 | 5 | 2 | 98014 | 47.6517 | -121.906 | 23103 | 1800 | 1800 | 0 | 1.0 | 3 | 1.75 | 0 | 1 | 3 | 7 | 1968 | 0 | 18163 | 1410 | 284000 |
| 10 | 10 | 2014 | 5 | 2 | 98023 | 47.3256 | -122.378 | 33151 | 3240 | 3240 | 0 | 2.0 | 3 | 2.50 | 0 | 3 | 3 | 10 | 1995 | 0 | 24967 | 4050 | 604000 |
| 11 | 11 | 2014 | 5 | 2 | 98024 | 47.5643 | -121.897 | 16215 | 1580 | 1580 | 0 | 1.0 | 3 | 2.25 | 0 | 1 | 4 | 7 | 1978 | 0 | 16215 | 1450 | 335000 |
| 12 | 12 | 2014 | 5 | 2 | 98027 | 47.4635 | -121.991 | 35100 | 1970 | 1970 | 0 | 2.0 | 3 | 2.25 | 0 | 1 | 4 | 9 | 1977 | 0 | 35100 | 2340 | 437500 |
| 13 | 13 | 2014 | 5 | 2 | 98027 | 47.4634 | -121.987 | 37277 | 2710 | 2710 | 0 | 2.0 | 4 | 2.75 | 0 | 1 | 3 | 9 | 2000 | 0 | 39299 | 2390 | 630000 |
| 14 | 14 | 2014 | 5 | 2 | 98029 | 47.5794 | -122.025 | 67518 | 2820 | 2820 | 0 | 2.0 | 5 | 2.50 | 0 | 1 | 3 | 8 | 1979 | 0 | 48351 | 2820 | 675000 |
- Split the house sales dataset into two tables. A training set, that we will use later to implement a training model to predict the house price, containing 60% of the data, and a testing set containing the rest of the data.
Delete the column
pricefrom the test set, to be able to predict it later:
In [2]:
Copied!
split_tuple = pricing.split_rows(0.60)
train_table = split_tuple[0]
testing_table = split_tuple[1]
test_table = testing_table.remove_columns(["price"]).shuffle_rows()
split_tuple = pricing.split_rows(0.60)
train_table = split_tuple[0]
testing_table = split_tuple[1]
test_table = testing_table.remove_columns(["price"]).shuffle_rows()
- Tag the
priceColumnas the target variable to be predicted. Use the new names of the fittedColumns as features, which will be used to make predictions based on the target variable.
In [3]:
Copied!
feature_columns = set(train_table.column_names) - set(["price", "id"])
tagged_train_table = train_table.tag_columns("price", feature_names=[
*feature_columns])
feature_columns = set(train_table.column_names) - set(["price", "id"])
tagged_train_table = train_table.tag_columns("price", feature_names=[
*feature_columns])
- Use
Decision Treeregressor as a model for the regression. Pass the "tagged_pricing" table to the fit function of the model:
In [4]:
Copied!
from safeds.ml.classical.regression import DecisionTree
model = DecisionTree()
fitted_model = model.fit(tagged_train_table)
from safeds.ml.classical.regression import DecisionTree
model = DecisionTree()
fitted_model = model.fit(tagged_train_table)
- Use the fitted decision tree regression model, that we trained on the training dataset to predict the price of a house in the test dataset.
In [5]:
Copied!
prediction = fitted_model.predict(
test_table
)
# For visualisation purposes we only print out the first 15 rows.
prediction.slice_rows(0,15)
prediction = fitted_model.predict(
test_table
)
# For visualisation purposes we only print out the first 15 rows.
prediction.slice_rows(0,15)
Out[5]:
| id | year | month | day | zipcode | latitude | longitude | sqft_lot | sqft_living | sqft_above | sqft_basement | floors | bedrooms | bathrooms | waterfront | view | condition | grade | year_built | year_renovated | sqft_lot_15nn | sqft_living_15nn | price | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 0 | 14656 | 2015 | 1 | 2 | 98058 | 47.4650 | -122.123 | 35171 | 2760 | 2760 | 0 | 2.0 | 3 | 2.50 | 0 | 1 | 3 | 9 | 1990 | 0 | 35171 | 2720 | 530000.0 |
| 1 | 19472 | 2015 | 4 | 10 | 98168 | 47.4717 | -122.323 | 9086 | 1480 | 1480 | 0 | 1.0 | 3 | 1.50 | 0 | 1 | 3 | 7 | 1962 | 0 | 9750 | 1540 | 248000.0 |
| 2 | 18494 | 2015 | 3 | 27 | 98053 | 47.6640 | -122.041 | 46538 | 3030 | 3030 | 0 | 2.0 | 3 | 2.50 | 0 | 1 | 3 | 10 | 1997 | 0 | 51450 | 3370 | 999000.0 |
| 3 | 14318 | 2014 | 12 | 22 | 98072 | 47.7434 | -122.106 | 19991 | 1990 | 1340 | 650 | 1.0 | 3 | 2.75 | 0 | 1 | 3 | 7 | 1977 | 0 | 9775 | 1750 | 450000.0 |
| 4 | 20542 | 2015 | 4 | 27 | 98028 | 47.7637 | -122.266 | 23030 | 1140 | 1140 | 0 | 1.0 | 2 | 1.00 | 0 | 1 | 3 | 8 | 1980 | 0 | 14260 | 1850 | 375000.0 |
| 5 | 21194 | 2015 | 5 | 5 | 98034 | 47.7190 | -122.173 | 8155 | 1770 | 1770 | 0 | 1.5 | 4 | 2.50 | 0 | 1 | 4 | 6 | 1970 | 1993 | 7360 | 1460 | 487585.0 |
| 6 | 14763 | 2015 | 1 | 6 | 98106 | 47.5352 | -122.361 | 4380 | 1230 | 1230 | 0 | 1.0 | 3 | 1.00 | 0 | 1 | 3 | 6 | 1947 | 0 | 6026 | 1525 | 343000.0 |
| 7 | 19223 | 2015 | 4 | 8 | 98033 | 47.6882 | -122.171 | 5107 | 1810 | 1810 | 0 | 2.0 | 3 | 2.25 | 0 | 1 | 3 | 8 | 1989 | 0 | 5454 | 1760 | 472500.0 |
| 8 | 14796 | 2015 | 1 | 7 | 98042 | 47.3774 | -122.160 | 13482 | 2980 | 1730 | 1250 | 1.0 | 5 | 2.75 | 0 | 1 | 4 | 8 | 1975 | 0 | 14800 | 2900 | 375000.0 |
| 9 | 21458 | 2015 | 5 | 8 | 98125 | 47.7153 | -122.284 | 5759 | 910 | 910 | 0 | 1.0 | 2 | 1.00 | 0 | 1 | 3 | 6 | 1951 | 0 | 7518 | 1520 | 371000.0 |
| 10 | 20634 | 2015 | 4 | 27 | 98144 | 47.5920 | -122.295 | 2268 | 1220 | 610 | 610 | 1.0 | 2 | 1.75 | 0 | 1 | 4 | 6 | 1909 | 0 | 1675 | 1240 | 850000.0 |
| 11 | 19912 | 2015 | 4 | 17 | 98092 | 47.3163 | -122.188 | 5250 | 1570 | 1570 | 0 | 1.0 | 3 | 2.00 | 0 | 1 | 3 | 7 | 1998 | 0 | 5250 | 1570 | 265000.0 |
| 12 | 21053 | 2015 | 5 | 3 | 98119 | 47.6501 | -122.370 | 5210 | 2010 | 1890 | 120 | 1.5 | 5 | 1.00 | 0 | 1 | 3 | 9 | 1927 | 0 | 5000 | 2330 | 1370000.0 |
| 13 | 17670 | 2015 | 3 | 16 | 98146 | 47.5088 | -122.371 | 7921 | 1430 | 1430 | 0 | 1.0 | 2 | 1.75 | 0 | 1 | 3 | 7 | 1983 | 0 | 8040 | 1290 | 361810.0 |
| 14 | 14826 | 2015 | 1 | 7 | 98148 | 47.4536 | -122.330 | 7582 | 1400 | 1400 | 0 | 1.0 | 3 | 1.50 | 0 | 1 | 3 | 7 | 1956 | 0 | 7872 | 1280 | 225000.0 |
- You can test the mean absolute error of that model with the initial testing_table as follows:
In [6]:
Copied!
tagged_test_table= testing_table.tag_columns("price", feature_names=[
*feature_columns
])
fitted_model.mean_absolute_error(tagged_test_table)
tagged_test_table= testing_table.tag_columns("price", feature_names=[
*feature_columns
])
fitted_model.mean_absolute_error(tagged_test_table)
Out[6]:
105232.47698091382