Spaces:
Runtime error
Runtime error
Upload app.py
Browse files
app.py
ADDED
@@ -0,0 +1,116 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# -*- coding: utf-8 -*-
|
2 |
+
import tensorflow_decision_forests as tfdf
|
3 |
+
import pandas as pd
|
4 |
+
import gradio as gr
|
5 |
+
import urllib
|
6 |
+
from tensorflow import keras
|
7 |
+
|
8 |
+
|
9 |
+
input_path = "https://archive.ics.uci.edu/ml/machine-learning-databases/census-income-mld/census-income"
|
10 |
+
input_column_header = "income_level"
|
11 |
+
|
12 |
+
#Load data
|
13 |
+
|
14 |
+
BASE_PATH = input_path
|
15 |
+
CSV_HEADER = [ l.decode("utf-8").split(":")[0].replace(" ", "_")
|
16 |
+
for l in urllib.request.urlopen(f"{BASE_PATH}.names")
|
17 |
+
if not l.startswith(b"|")][2:]
|
18 |
+
|
19 |
+
CSV_HEADER.append(input_column_header)
|
20 |
+
|
21 |
+
train_data = pd.read_csv(f"{BASE_PATH}.data.gz", header=None, names=CSV_HEADER)
|
22 |
+
test_data = pd.read_csv(f"{BASE_PATH}.test.gz", header=None, names=CSV_HEADER)
|
23 |
+
|
24 |
+
#subset data
|
25 |
+
train_data = train_data.loc[:, ["education", "sex", "capital_gains", "capital_losses", "income_level"]]
|
26 |
+
test_data = test_data.loc[:, ["education", "sex", "capital_gains", "capital_losses", "income_level"]]
|
27 |
+
|
28 |
+
def encode_df(df):
|
29 |
+
sex_mapping = {" Male": 0, " Female": 1}
|
30 |
+
df = df.replace({"sex": sex_mapping})
|
31 |
+
education_mapping = {" High school graduate": 1, " Some college but no degree": 2,
|
32 |
+
" 10th grade": 3, " Children": 4, " Bachelors degree(BA AB BS)": 5,
|
33 |
+
" Masters degree(MA MS MEng MEd MSW MBA)": 6, " Less than 1st grade": 7,
|
34 |
+
" Associates degree-academic program": 8, " 7th and 8th grade": 9,
|
35 |
+
" 12th grade no diploma": 10, " Associates degree-occup /vocational": 11,
|
36 |
+
" Prof school degree (MD DDS DVM LLB JD)": 12, " 5th or 6th grade": 13,
|
37 |
+
" 11th grade": 14, " Doctorate degree(PhD EdD)": 15, " 9th grade": 16,
|
38 |
+
" 1st 2nd 3rd or 4th grade": 17}
|
39 |
+
df = df.replace({"education": education_mapping})
|
40 |
+
income_mapping = {' - 50000.': 0, ' 50000+.': 1}
|
41 |
+
df = df.replace({"income_level": income_mapping})
|
42 |
+
return df
|
43 |
+
|
44 |
+
train_data = encode_df(train_data)
|
45 |
+
test_data = encode_df(test_data)
|
46 |
+
|
47 |
+
feature_a = tfdf.keras.FeatureUsage(name="education", semantic=tfdf.keras.FeatureSemantic.CATEGORICAL)
|
48 |
+
feature_b = tfdf.keras.FeatureUsage(name="sex", semantic=tfdf.keras.FeatureSemantic.CATEGORICAL)
|
49 |
+
feature_c = tfdf.keras.FeatureUsage(name="capital_gains", semantic=tfdf.keras.FeatureSemantic.CATEGORICAL)
|
50 |
+
feature_d = tfdf.keras.FeatureUsage(name="capital_losses", semantic=tfdf.keras.FeatureSemantic.CATEGORICAL)
|
51 |
+
|
52 |
+
# Convert the dataset into a TensorFlow dataset.
|
53 |
+
train_ds = tfdf.keras.pd_dataframe_to_tf_dataset(train_data, label="income_level")
|
54 |
+
test_ds = tfdf.keras.pd_dataframe_to_tf_dataset(test_data, label="income_level")
|
55 |
+
|
56 |
+
# Train a GB Trees model
|
57 |
+
model = tfdf.keras.GradientBoostedTreesModel(
|
58 |
+
features = [feature_a, feature_b, feature_c, feature_d],
|
59 |
+
exclude_non_specified_features = True,
|
60 |
+
growing_strategy = "BEST_FIRST_GLOBAL",
|
61 |
+
num_trees = 350,
|
62 |
+
max_depth = 7,
|
63 |
+
min_examples = 6,
|
64 |
+
subsample = 0.65,
|
65 |
+
sampling_method = "GOSS",
|
66 |
+
validation_ratio = 0.1,
|
67 |
+
task = tfdf.keras.Task.CLASSIFICATION,
|
68 |
+
loss = "DEFAULT",
|
69 |
+
verbose=0)
|
70 |
+
|
71 |
+
model.compile(metrics=[keras.metrics.BinaryAccuracy(name="accuracy")])
|
72 |
+
model.fit(train_ds)
|
73 |
+
model.evaluate(test_ds)
|
74 |
+
|
75 |
+
#prepare user input for the model
|
76 |
+
def process_inputs(education, sex, capital_gains, capital_losses):
|
77 |
+
df = pd.DataFrame.from_dict(
|
78 |
+
{
|
79 |
+
"education": [edu_in],
|
80 |
+
"sex": [sex_in],
|
81 |
+
"capital_gains": [cap_gains_in],
|
82 |
+
"capital_losses": [cap_losses_in]
|
83 |
+
}
|
84 |
+
)
|
85 |
+
df = encode_df(df)
|
86 |
+
|
87 |
+
feature_a = tfdf.keras.FeatureUsage(name="education", semantic=tfdf.keras.FeatureSemantic.CATEGORICAL)
|
88 |
+
feature_b = tfdf.keras.FeatureUsage(name="sex", semantic=tfdf.keras.FeatureSemantic.CATEGORICAL)
|
89 |
+
feature_c = tfdf.keras.FeatureUsage(name="capital_gains", semantic=tfdf.keras.FeatureSemantic.CATEGORICAL)
|
90 |
+
feature_d = tfdf.keras.FeatureUsage(name="capital_losses", semantic=tfdf.keras.FeatureSemantic.CATEGORICAL)
|
91 |
+
|
92 |
+
df = tfdf.keras.pd_dataframe_to_tf_dataset(df)
|
93 |
+
|
94 |
+
pred = model.predict(df)
|
95 |
+
if pred > .5:
|
96 |
+
pred_bi = 1
|
97 |
+
return {"> $50,000": pred_bi}
|
98 |
+
elif pred <=.5:
|
99 |
+
pred_bi = 0
|
100 |
+
return {"<= $50,000": pred_bi}
|
101 |
+
|
102 |
+
iface = gr.Interface(
|
103 |
+
process_inputs,
|
104 |
+
[
|
105 |
+
gr.inputs.Dropdown([" 1st 2nd 3rd or 4th grade", " High school graduate",
|
106 |
+
" Bachelors degree(BA AB BS)", " Masters degree(MA MS MEng MEd MSW MBA)",
|
107 |
+
" Prof school degree (MD DDS DVM LLB JD)",
|
108 |
+
" Doctorate degree(PhD EdD)"], type="index", label="education"),
|
109 |
+
gr.inputs.Radio([" Male", " Female"], label="sex", type="index"),
|
110 |
+
gr.inputs.Slider(minimum = 0, maximum = 99999, label="capital_gains"),
|
111 |
+
gr.inputs.Slider(minimum = 0, maximum = 4608, label="capital_losses")
|
112 |
+
],
|
113 |
+
gr.outputs.Label(num_top_classes=2),
|
114 |
+
live=True,
|
115 |
+
analytics_enabled=False
|
116 |
+
)
|