Spaces:
Runtime error
Runtime error
File size: 5,524 Bytes
56f6887 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 |
import pandas as pd
import numpy as np
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import LabelEncoder, MinMaxScaler
import math
import matplotlib.pyplot as plt, joblib
# from sklearn.linear_model import LinearRegression
from sklearn.ensemble import RandomForestRegressor
# from sklearn.tree import DecisionTreeRegressor
# from catboost import CatBoostRegressor
import warnings, random
from sklearn.metrics import mean_absolute_error as mae
from sklearn.metrics import mean_squared_error as mse
# from sklearn import tree
# from sklearn.svm import SVR
# from sklearn.ensemble import VotingRegressor
import os
warnings.filterwarnings("ignore")
features = [
"batting_team",
# "bowling_team",
# "balls",
# "runs",
# "wickets",
"wkt_last_5_overs",
# "runrate_last_5_overs",
"current_RR",
# "average",
"balls_left",
"wkts_left",
# "required_RR",
# "projected_score_more",
# "min_score_more",
# "max_score_more",
# "projected_avg_score_more",
"runrate_last_5_overs-current_RR",
]
target = "deviation_from_projected"
# evaluate
def evaluate(model, featuresdf, x_test, fname):
predictdf = featuresdf.loc[x_test.index].copy()
# print(predictdf.columns)
predictdf["h_deviation_from_projected"] = model.predict(
featuresdf.loc[x_test.index][features]
)
predictdf["error"] = (
predictdf["h_deviation_from_projected"] - predictdf["deviation_from_projected"]
)
predictdf["abs_error"] = predictdf["error"].abs()
ax = plt.gca()
plt.plot(predictdf.groupby("balls").aggregate({"abs_error": "mean"}))
plt.legend("Abs deviation")
# ax.set_ylim([-50, 50])
plt.title(type(model).__name__)
plt.xlabel("Balls on which prediction was made")
plt.ylabel("Mean Abs Prediction error")
plt.savefig("result/" + fname + ".png")
plt.clf()
predictdf.sample(frac=0.0001).to_csv("result/" + fname + "_sample.csv")
# fig = plt.figure(figsize=(25, 20))
# tree.plot_tree(model)
# fig.savefig(fname + ".png")
# plt.clf()batting_teamsort_values("overs", ascending=False).to_string(index=False))
def train_test_split_matchid(df, matchids, split=0.2):
unique_match_ids = set(matchids)
print(f"{len(unique_match_ids)=}")
testids = random.sample(unique_match_ids, int(len(unique_match_ids) * split))
trainids = list(unique_match_ids.difference(testids))
return (
df[features][df.matchid.isin(trainids)],
df[features][df.matchid.isin(testids)],
df[target][df.matchid.isin(trainids)],
df[target][df.matchid.isin(testids)],
)
def encode_teams(series):
encoder = LabelEncoder()
encoder.fit(series)
np.save("model/team.npy", encoder.classes_)
def transform_teams(series):
encoder = LabelEncoder()
encoder.classes_ = np.load("model/team.npy", allow_pickle=True)
return encoder.transform(np.array(series).reshape(-1, 1)).reshape(-1)
def plot_feature_importance(f, imp, fname):
importance = (
pd.DataFrame(
zip(*[f, imp]),
columns=["feature", "importance"],
)
.sort_values("importance", ascending=False)
.set_index("feature")
)
importance["importance"] = importance["importance"] / importance["importance"].sum()
fig, ax = plt.subplots()
importance.plot.bar(ax=ax)
ax.bar_label(ax.containers[0], labels=f, rotation=90, label_type="center")
ax.set_xticks([])
ax.set_title("Feature importances for predicted score " + fname)
ax.set_ylabel("Significance")
ax.set_xlabel("Features")
plt.savefig("result/" + fname + "featureimp.png")
plt.clf()
def train(fname, max_depth=-1):
print("training on", fname, "...")
featuresdf = pd.read_feather(fname)
featuresdf = featuresdf[featuresdf["inning"] == 2]
encode_teams(
featuresdf["batting_team"].to_list() + featuresdf["bowling_team"].to_list()
)
featuresdf["batting_team"] = transform_teams(featuresdf["batting_team"])
featuresdf["bowling_team"] = transform_teams(featuresdf["bowling_team"])
x_train, x_test, y_train, y_test = train_test_split_matchid(
featuresdf, featuresdf["matchid"], 0.2
)
print(f"{len(x_train)=} {len(x_test)=}")
model = RandomForestRegressor(max_depth=8)
model.fit(x_train, y_train)
# for xgb
# plot_feature_importance(
# model.get_booster().get_score(importance_type="gain").keys(),
# model.get_booster().get_score(importance_type="gain").values(),
# fname,
# )
# for rf
plot_feature_importance(
features,
np.std([tree.feature_importances_ for tree in model.estimators_], axis=0),
os.path.basename(fname),
)
print("Depth:", [e.tree_.max_depth for e in model.estimators_])
# for dt
# plot_feature_importance(
# features,
# model.feature_importances_,
# fname,
# )
# print(model.tree_.max_depth)
# print(f"{model.score(x_train, y_train)=}, {model.score(x_test, y_test)=}")
print(
f"{mse(model.predict(x_train), y_train, squared=False)=}, {mse(model.predict(x_test), y_test, squared=False)=}"
)
evaluate(model, featuresdf, x_test, os.path.basename(fname))
model.fit(featuresdf[features], featuresdf[target])
joblib.dump(model, f"model/{os.path.basename(fname)}.joblib")
return model
if __name__ == "__main__":
train("data/t20features.feather")
train("data/odifeatures.feather")
|