Datasets:
Update diamonds.py
Browse files- diamonds.py +26 -13
diamonds.py
CHANGED
@@ -69,8 +69,20 @@ features_types_per_config = {
|
|
69 |
"observation_point_on_axis_y": datasets.Value("float32"),
|
70 |
"observation_point_on_axis_z": datasets.Value("float32"),
|
71 |
"cut": datasets.ClassLabel(num_classes=5, names=("Fair", "Good", "Very Good", "Premium", "Ideal"))
|
72 |
-
}
|
73 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
74 |
}
|
75 |
features_per_config = {k: datasets.Features(features_types_per_config[k]) for k in features_types_per_config}
|
76 |
|
@@ -85,10 +97,9 @@ class Diamond(datasets.GeneratorBasedBuilder):
|
|
85 |
# dataset versions
|
86 |
DEFAULT_CONFIG = "cut"
|
87 |
BUILDER_CONFIGS = [
|
88 |
-
DiamondConfig(name="encoding",
|
89 |
-
|
90 |
-
DiamondConfig(name="
|
91 |
-
description="5-ary classification, predict the cut quality of the diamond."),
|
92 |
]
|
93 |
|
94 |
|
@@ -118,18 +129,20 @@ class Diamond(datasets.GeneratorBasedBuilder):
|
|
118 |
yield row_id, data_row
|
119 |
|
120 |
def preprocess(self, data: pandas.DataFrame, config: str = "cut") -> pandas.DataFrame:
|
121 |
-
data
|
122 |
-
data
|
123 |
-
data
|
124 |
-
data
|
125 |
|
126 |
for feature in _ENCODING_DICS:
|
127 |
encoding_function = partial(self.encode, feature)
|
128 |
-
data
|
129 |
|
130 |
data.columns = _BASE_FEATURE_NAMES
|
131 |
-
data = data.drop_duplicates(subset=["carat", "color", "clarity", "depth", "table",
|
132 |
-
|
|
|
|
|
133 |
|
134 |
|
135 |
return data[list(features_types_per_config["cut"].keys())]
|
|
|
69 |
"observation_point_on_axis_y": datasets.Value("float32"),
|
70 |
"observation_point_on_axis_z": datasets.Value("float32"),
|
71 |
"cut": datasets.ClassLabel(num_classes=5, names=("Fair", "Good", "Very Good", "Premium", "Ideal"))
|
72 |
+
},
|
73 |
+
|
74 |
+
"cut_binary": {
|
75 |
+
"carat": datasets.Value("float32"),
|
76 |
+
"color": datasets.Value("string"),
|
77 |
+
"clarity": datasets.Value("float32"),
|
78 |
+
"depth": datasets.Value("float32"),
|
79 |
+
"table": datasets.Value("float32"),
|
80 |
+
"price": datasets.Value("float32"),
|
81 |
+
"observation_point_on_axis_x": datasets.Value("float32"),
|
82 |
+
"observation_point_on_axis_y": datasets.Value("float32"),
|
83 |
+
"observation_point_on_axis_z": datasets.Value("float32"),
|
84 |
+
"cut": datasets.ClassLabel(num_classes=2, names=("no", "yes"))
|
85 |
+
},
|
86 |
}
|
87 |
features_per_config = {k: datasets.Features(features_types_per_config[k]) for k in features_types_per_config}
|
88 |
|
|
|
97 |
# dataset versions
|
98 |
DEFAULT_CONFIG = "cut"
|
99 |
BUILDER_CONFIGS = [
|
100 |
+
DiamondConfig(name="encoding", description="Encoding dictionaries for discrete features."),
|
101 |
+
DiamondConfig(name="cut", description="5-ary classification, predict the cut quality of the diamond."),
|
102 |
+
DiamondConfig(name="cut_binary", description="Binary classification."),
|
|
|
103 |
]
|
104 |
|
105 |
|
|
|
129 |
yield row_id, data_row
|
130 |
|
131 |
def preprocess(self, data: pandas.DataFrame, config: str = "cut") -> pandas.DataFrame:
|
132 |
+
data["clarity"] = data.clarity.apply(lambda x: x.replace("b", "").replace("'", ""))
|
133 |
+
data["cut"] = data.cut.apply(lambda x: x.replace("b", "").replace("'", ""))
|
134 |
+
data["color"] = data.color.astype(str)
|
135 |
+
data["color"] = data.color.apply(lambda x: x[2]).replace("\"", "")
|
136 |
|
137 |
for feature in _ENCODING_DICS:
|
138 |
encoding_function = partial(self.encode, feature)
|
139 |
+
data[feature] = data[feature].apply(encoding_function)
|
140 |
|
141 |
data.columns = _BASE_FEATURE_NAMES
|
142 |
+
data = data.drop_duplicates(subset=["carat", "color", "clarity", "depth", "table", "price", "cut"])
|
143 |
+
|
144 |
+
if self.config.name == "cut_binary":
|
145 |
+
data.cut = data.cut.apply(lambda x: 0 if x <= 2 else 1)
|
146 |
|
147 |
|
148 |
return data[list(features_types_per_config["cut"].keys())]
|