Skip to content

Instantly share code, notes, and snippets.

View makispl's full-sized avatar

Plegas Gerasimos makispl

25 followers · 35 following
View GitHub Profile
@makispl
makispl / bayesian_gm.py
Last active August 31, 2021 11:08
# find the optimum number of clusters
bgm = BayesianGaussianMixture(n_components=10, n_init=7, max_iter=1000)
bgm.fit(pca_scores)
np.round(bgm.weights_, 2)
@makispl
makispl / pca4clustering.py
Last active August 31, 2021 13:22
# read in the training data
plays_df = pd.read_csv('../data/interim/plays_17_18_19_pre_proc_train.csv',
converters={'GAME_ID': lambda x: str(x)})
# switch to the for-normalisation-features
data_stnd = data.copy()
# instantiate, fit, transform scaler
scaler = MinMaxScaler()
data_stnd = scaler.fit_transform(data_stnd)
@makispl
makispl / kmeans.py
Last active August 31, 2021 10:21
# Instantiate a KMeans model with 4 clusters, fit and predict cluster indices
kmeans_pca = KMeans(n_clusters=4, init='random', random_state=1)
kmeans_pca.fit_predict(pca_scores)
plays_km_df['km_cluster'] = kmeans_pca.labels_
# concat plays_km_df with the pca components
plays_pca_km_df = pd.concat([plays_km_df.reset_index(drop=True), pd.DataFrame(
data=pca_scores, columns=['pca_1', 'pca_2', 'pca_3', 'pca_4'])], axis=1)
# visualize clusters
@makispl
makispl / locate_elbow.py
Last active August 31, 2021 09:43
# Switch to a new dataframe instance
# for the k-memans implementation
plays_km_df = plays_df.copy()
# Calculate the wcss
max_clusters = 11
wcss = list()
for k in range(1, max_clusters):
kmeans = KMeans(n_clusters=k, init='random', random_state=1)
@makispl
makispl / plays_all_descriptive_stats.tsv
Created August 31, 2021 08:53
OFF_RATING AST_PCT AST_TOV TM_TOV_PCT EFG_PCT TS_PCT POSS
count 62893.000000 62893.000000 62893.000000 62893.000000 62893.000000 62893.000000 62893.000000
mean 104.810268 0.138187 1.037766 10.242790 0.487106 0.519022 47.786987
std 23.251576 0.140800 1.579515 12.341289 0.281903 0.270811 21.401744
min 0.000000 0.000000 0.000000 0.000000 0.000000 0.000000 0.000000
25% 93.800000 0.000000 0.000000 0.000000 0.333000 0.375000 33.000000
50% 106.300000 0.111000 0.330000 8.300000 0.500000 0.532000 50.000000
75% 118.200000 0.208000 1.500000 15.400000 0.667000 0.680000 65.000000
max 350.000000 1.000000 17.000000 100.000000 1.500000 1.500000 121.000000
@makispl
makispl / plays_95th_descriptive_stats.tsv
Last active August 31, 2021 09:04
OFF_RATING AST_PCT AST_TOV TM_TOV_PCT EFG_PCT TS_PCT POSS
count 3056.000000 3056.000000 3056.000000 3056.000000 3056.000000 3056.000000 3056.000000
mean 104.781872 0.527235 2.929741 9.501309 0.472252 0.509817 48.632199
std 23.812656 0.142249 2.974245 8.700416 0.274866 0.267038 25.562033
min 22.200000 0.406000 0.000000 0.000000 0.000000 0.000000 1.000000
25% 93.000000 0.440000 0.000000 0.000000 0.333000 0.383000 28.000000
50% 105.900000 0.500000 2.310000 8.700000 0.500000 0.532000 53.000000
75% 118.200000 0.545000 4.000000 14.300000 0.625000 0.666000 70.000000
max 250.000000 1.000000 17.000000 66.700000 1.500000 1.500000 121.000000
@makispl
makispl / features_explained.tsv
Last active August 31, 2021 08:18
Column Definition Group
OFF_RATING Points scored per 100 possessions group_1
AST_PCT Ratio of player assists to teammate field goals made group_1
AST_TOV Ratio of assists to turnovers group_1
TM_TOV_PCT Ratio of turnovers to possessions group_1
EFG_PCT Weighted field goal percentage with 3-point field goals weighted 1.5 times 2-point field goals group_1
TS_PCT Shooting percentage accounting for free throws and three-point field goals group_1
POSS Possession group_1
MIN Minutes played group_2
AST_RATIO Assists per 100 possessions group_2
@makispl
makispl / plays_df.csv
Last active August 31, 2021 07:22
'plays' Dataset Sample
We can make this file beautiful and searchable if this error is corrected: It looks like row 5 should actually have 33 columns, instead of 17. in line 4.
GAME_ID,GAME_DATE,TEAM_ID,TEAM_ABBREVIATION,TEAM_CITY,PLAYER_ID,PLAYER_NAME,NICKNAME,START_POSITION,COMMENT,MIN,E_OFF_RATING,OFF_RATING,E_DEF_RATING,DEF_RATING,E_NET_RATING,NET_RATING,AST_PCT,AST_TOV,AST_RATIO,OREB_PCT,DREB_PCT,REB_PCT,TM_TOV_PCT,EFG_PCT,TS_PCT,USG_PCT,E_USG_PCT,E_PACE,PACE,PACE_PER40,POSS,PIE
0021700002,2017-10-17,1610612745,HOU,Houston,2772,Trevor Ariza,Trevor,F,,37:39,107.3,111.3,121.4,121.3,-14.1,-10.0,0.152,5.0,33.3,0.023,0.156,0.079,6.7,0.444,0.444,0.109,0.111,103.83,101.99,84.99,80,0.063
0021700002,2017-10-17,1610612745,HOU,Houston,201583,Ryan Anderson,Ryan,F,,33:17,117.8,127.4,119.4,119.7,-1.7,7.7,0.031,0.0,7.7,0.079,0.2,0.127,0.0,0.542,0.542,0.141,0.141,108.25,103.84,86.53,73,0.064
0021700002,2017-10-17,1610612745,HOU,Houston,203991,Clint Capela,Clint,C,,18:10,95.4,95.1,153.0,155.0,-57.6,-59.9,0.0,0.0,0.0,0.05,0.25,0.125,16.7,0.6,0.575,0.267,0.273,107.54,107.01,89.17,41,0.092
0021700002,2017-10-17,1610612745,HOU,Houston,201935,James Harden,James,G,,36:24,117.6,126.0,126.2,124.7,-8.6,
@makispl
makispl / mature_whiskies_price.py
Created February 22, 2021 19:54
display(df_num[df_num['age'] > 30]['price'].sort_values(ascending=False).head(10))
@makispl
makispl / logistic_regression.py
Created February 22, 2021 19:49
# Switch to a copy of the labeled dataframe
df_no_nuls_2 = df_no_nuls.copy()
# Randomise the df
shuffled_rows = np.random.permutation(df_no_nuls_2.index)
df_no_nuls_2 = df_no_nuls_2.loc[shuffled_rows]
# Split to train and test datasets
train = df_no_nuls_2.iloc[:int(df_no_nuls_2.shape[0]*0.8)].copy()
test = df_no_nuls_2.iloc[int(df_no_nuls_2.shape[0]*0.8):].copy().reset_index()