 import gzip
import nltk
import numpy as np
import mlflow
import pandas as pd
import shutil
import time
import plotly.graph_objects as go
import plotly.express as px
from matplotlib import pyplot as plt
from pandas_profiling import ProfileReport
from os.path import exists
from sklearn.datasets import make_classification
from sklearn.ensemble import RandomForestClassifier
from sklearn.feature_extraction.text import TfidfVectorizer
from sklearn.linear_model import LogisticRegression
from sklearn.metrics import accuracy_score, recall_score, precision_score, f1_score, classification_report, confusion_matrix
from sklearn.model_selection import train_test_split,cross_val_score
from sklearn.svm import SVC
from urllib import request
from urllib.parse import urlparse

 def download_file(url):
  out_file = 'data/reviews_Video_Games_5.json.gz'
  file_exists = exists(out_file)
  if not file_exists:
    try:
      r1 = request.urlretrieve(url=url, filename=out_file)
      print(f'downloaded at {out_file}')
      return 1
    except Exception as e:
      print(e)
      return -1
  else:
    print('File already exists')
    return 0

def parse(path):
  g = gzip.open(path, 'rb')
  for l in g:
    yield eval(l)

def getDF(path):
  i = 0
  df = {}
  for d in parse(path):
    df[i] = d
    i += 1
  return pd.DataFrame.from_dict(df, orient='index')

 download_file('http://snap.stanford.edu/data/amazon/productGraph/categoryFiles/reviews_Video_Games_5.json.gz')

File already exists

0

 videogame_review_data_raw = getDF('data/reviews_Video_Games_5.json.gz')

 videogame_review_data_raw.head(10)

 videogame_review_data_raw.info()

<class 'pandas.core.frame.DataFrame'>
Int64Index: 231780 entries, 0 to 231779
Data columns (total 9 columns):
 #   Column          Non-Null Count   Dtype  
---  ------          --------------   -----  
 0   reviewerID      231780 non-null  object 
 1   asin            231780 non-null  object 
 2   reviewerName    228967 non-null  object 
 3   helpful         231780 non-null  object 
 4   reviewText      231780 non-null  object 
 5   overall         231780 non-null  float64
 6   summary         231780 non-null  object 
 7   unixReviewTime  231780 non-null  int64  
 8   reviewTime      231780 non-null  object 
dtypes: float64(1), int64(1), object(7)
memory usage: 17.7+ MB

 videogame_review_data_raw.count()

reviewerID        231780
asin              231780
reviewerName      228967
helpful           231780
reviewText        231780
overall           231780
summary           231780
unixReviewTime    231780
reviewTime        231780
dtype: int64

 videogame_review_data_raw.isna().sum()

reviewerID           0
asin                 0
reviewerName      2813
helpful              0
reviewText           0
overall              0
summary              0
unixReviewTime       0
reviewTime           0
dtype: int64

 def generate_profile_report(data, title):
    profile_report = ProfileReport(data, minimal=True, title=title)
    profile_report.to_file(title)
    profile_report.to_notebook_iframe()

 generate_profile_report(data=videogame_review_data_raw, title="Video game Dataset Raw")

Summarize dataset: 100%|██████████| 18/18 [00:03<00:00,  5.38it/s, Completed]
Generate report structure: 100%|██████████| 1/1 [00:02<00:00,  2.26s/it]
Render HTML: 100%|██████████| 1/1 [00:00<00:00,  8.51it/s]
C:\Users\squal\anaconda3\envs\modulo_6\lib\site-packages\pandas_profiling\profile_report.py:261: UserWarning:

Extension  not supported. For now we assume .html was intended. To remove this warning, please use .html or .json.

Export report to file: 100%|██████████| 1/1 [00:00<00:00, 251.97it/s]

 labels = ['5.0 Overall','4.0 Overall','3.0 Overall','2.0 Overall','1.0 Overall']
values = videogame_review_data_raw.overall.value_counts(normalize=True)

fig = go.Figure(data=[go.Pie(labels=labels, values=values)])
fig.update_traces(textposition='inside', textinfo='percent+label')
fig.update_layout(
    title='Distribución de Categorías',
    width=600,
    height=600,
    paper_bgcolor="LightSteelBlue",
)
fig.show(renderer='notebook')

 df = px.data.tips()
fig = px.histogram(videogame_review_data_raw, x="overall")
fig.show(renderer='notebook')

 duplicated_reviews=videogame_review_data_raw[videogame_review_data_raw.duplicated(['reviewText','overall'], keep=False)]
print(f'Reviews duplicados: {len(duplicated_reviews), len(videogame_review_data_raw)}')

Reviews duplicados: (126, 231780)

 videogame_review_data_raw.drop_duplicates(subset=['reviewText','overall'], inplace=True)
print(f'Duplicated reviews: {len(duplicated_reviews), len(videogame_review_data_raw)}')

Duplicated reviews: (126, 231694)

 videogame_review_data = pd.DataFrame({'review_text': pd.Series(dtype='str'),
                   'sentiment': pd.Series(dtype='int')})

videogame_review_data["review_text"] = videogame_review_data_raw["reviewText"].astype("string")
videogame_review_data["sentiment"] = videogame_review_data_raw["overall"].astype("int")
videogame_review_data.dtypes

review_text    string
sentiment       int32
dtype: object

 videogame_review_data.head(5)

 videogame_review_data=videogame_review_data[videogame_review_data['sentiment'] != 3]
print(f"Total de registros: {len(videogame_review_data)}")

Total de registros: 203420

 print(videogame_review_data['sentiment'].value_counts())
cat_5_count, cat_4_count, cat_2_count, cat_1_count  = videogame_review_data['sentiment'].value_counts()

5    120120
4     54789
1     14848
2     13663
Name: sentiment, dtype: int64

 cat_1_data = videogame_review_data[videogame_review_data['sentiment'] == 1]
cat_2_data = videogame_review_data[videogame_review_data['sentiment'] == 2]
cat_4_data = videogame_review_data[videogame_review_data['sentiment'] == 4]
cat_5_data = videogame_review_data[videogame_review_data['sentiment'] == 5]

 videogame_review_balanced_data = pd.concat([cat_1_data.sample(cat_1_count, replace=True),cat_2_data.sample(cat_2_count, replace=True),cat_4_data.sample(cat_1_count, replace=True),cat_5_data.sample(cat_1_count, replace=True)])
videogame_review_balanced_data.shape

(55837, 2)

  def label_sentiment(row):
    if int(row['sentiment']) < 3:
        return 'neg'
    else:
        return 'pos'

 videogame_review_balanced_data['sentiment_label'] = videogame_review_balanced_data.apply(lambda row: label_sentiment(row), axis=1)
videogame_review_balanced_data.head()

 x_train, x_test, y_train, y_test = train_test_split(
    videogame_review_balanced_data['review_text'],
    videogame_review_balanced_data['sentiment_label'],
    train_size=0.80,
    test_size=0.20,
    random_state=42,
    shuffle=True,
    stratify=videogame_review_balanced_data['sentiment_label']
)

 videogame_review_balanced_data['sentiment_label']

138127    neg
169555    neg
109323    neg
64521     neg
187417    neg
         ... 
114435    pos
96196     pos
59329     pos
10912     pos
30011     pos
Name: sentiment_label, Length: 55837, dtype: object

 def train_vectorizer(max_features=2500, data=None):
    cv = TfidfVectorizer(
        max_df=0.95,  # Elimina el 5% de los tokens más comunes
        min_df=5,  # No tiene en cuenta aquellos tokens que aparezcan menos de 5 veces
        max_features=max_features,  # Máximo tokens en el vocabulario
        strip_accents='ascii',  # Elimina acentos durante la normalización
        analyzer = 'word',
        lowercase=True, #convertimos a minusculas
        stop_words='english', #aplicamos lista de stop_words
        ngram_range=(2, 3)  # Identificará y utilizará como features bigramas y trigramas
    )
    cv.fit(data)
    return cv

 vectorizer_small = train_vectorizer(max_features=2500,data=x_train)
vectorizer_medium = train_vectorizer(max_features=10000,data=x_train)
vectorizer_large = train_vectorizer(max_features=50000,data=x_train)

 print(list(vectorizer_small.vocabulary_.items())[:20])
print(list(vectorizer_small.idf_)[:20])
print(len(vectorizer_small.vocabulary_))

[('minutes game', 1599), ('game games', 710), ('going game', 1048), ('game just', 743), ('touch screen', 2293), ('don care', 389), ('rated game', 1927), ('avoid game', 105), ('30 40', 24), ('good game', 1059), ('game want', 930), ('games just', 986), ('just trying', 1318), ('great games', 1136), ('games like', 988), ('final fantasy', 531), ('fantasy series', 502), ('right away', 2017), ('game given', 715), ('things like', 2232)]
[6.696422120249909, 6.772583481215473, 7.207247744015899, 7.218421044614025, 6.234786740674689, 5.917097243448911, 7.062666515204792, 6.772583481215473, 7.1531805227456235, 6.256018960780464, 6.9534672232398, 7.122089935675592, 7.207247744015899, 7.0723284261165285, 6.936372789880499, 6.308894712828412, 7.185268837297124, 7.229720599867958, 6.970858965951669, 7.122089935675592]
2500

 print(list(vectorizer_medium.vocabulary_.items())[:20])
print(list(vectorizer_medium.idf_)[:20])
print(len(vectorizer_medium.vocabulary_))

[('minutes game', 6186), ('game games', 2947), ('tries hard', 9164), ('going game', 3961), ('game just', 3057), ('star fox', 8454), ('touch screen', 9141), ('control stick', 1219), ('don care', 1764), ('rated game', 7570), ('just stupid', 5050), ('avoid game', 411), ('game coming', 2739), ('30 40', 72), ('40 game', 112), ('game systems', 3464), ('actually makes', 234), ('makes good', 6020), ('good game', 4024), ('game want', 3538)]
[8.488181589477964, 7.755813695764737, 6.696422120249909, 7.795034408918018, 6.772583481215473, 7.207247744015899, 7.389569300809854, 7.646614403799745, 7.969387796062796, 7.218421044614025, 8.15170935285675, 8.241321511546438, 6.234786740674689, 7.945857298652602, 8.09613950170194, 5.917097243448911, 7.062666515204792, 6.772583481215473, 7.1531805227456235, 7.350348587656573]
10000

 print(list(vectorizer_large.vocabulary_.items())[:20])
print(list(vectorizer_large.idf_)[:20])
print(len(vectorizer_large.vocabulary_))

[('minutes game', 29711), ('game games', 15262), ('tries hard', 45385), ('idea supposed', 22645), ('going game', 18883), ('game just', 15595), ('star fox', 41979), ('touch screen', 45249), ('control stick', 7319), ('stick don', 42279), ('don care', 9898), ('rated game', 37100), ('just stupid', 24368), ('avoid game', 2262), ('game coming', 14693), ('games 30', 17443), ('30 40', 372), ('40 game', 578), ('game systems', 16807), ('saying games', 39745)]
[9.309162141547795, 8.448960876324684, 8.488181589477964, 9.404472321352118, 9.222150764558164, 9.142108056884627, 9.404472321352118, 9.068000084730906, 9.404472321352118, 9.222150764558164, 9.404472321352118, 8.571563198417014, 7.755813695764737, 6.696422120249909, 9.309162141547795, 9.404472321352118, 9.309162141547795, 8.711325140792173, 7.795034408918018, 8.529003583998218]
50000

 x_train_small_transformed = vectorizer_small.transform(x_train)
x_test_small_transformed = vectorizer_small.transform(x_test)

x_train_medium_transformed = vectorizer_medium.transform(x_train)
x_test_medium_transformed = vectorizer_medium.transform(x_test)

x_train_large_transformed = vectorizer_large.transform(x_train)
x_test_large_transformed = vectorizer_large.transform(x_test)

 class experiment_model_log:
    model=None
    name=None
    classification_report=None

    def __init__(self, model, name, classification_report):
        self.model = model
        self.name = name
        self.classification_report = classification_report

 def train_logistic_regression_model(x_train, x_test, y_train, y_test, max_iter=1000, c_param=1, solver='lbfgs',run_name=''):
    with mlflow.start_run(run_name=run_name):
        #Definición de modelo con algoritmo
        model = LogisticRegression(C=c_param, solver=solver, max_iter=max_iter)

        #Entrenamiento del modelo
        model.fit(x_train, y_train)

        #Obtenemos predicciones
        #train_predict = model.predict(x_train)
        test_predict = model.predict(x_test)
    
        #Obtenemos métricas
        print(f"Test accuracy: {accuracy_score(y_test, test_predict)}")
        classification_report_model=classification_report(y_test, test_predict,output_dict=True)

        #Registro de Parametros en MLFlow
        mlflow.log_param("C", c_param)
        mlflow.log_param("max_iter", max_iter)
        mlflow.log_param("solver", solver)

        #Registro de Metricas en MLFlow
        mlflow.log_metric("Accuracy", classification_report_model.get('accuracy'))
        mlflow.log_metric("Precision", classification_report_model.get('macro avg').get('precision'))
        mlflow.log_metric("Recall", classification_report_model.get('macro avg').get('recall'))
        mlflow.log_metric("f-1 score", classification_report_model.get('macro avg').get('f1-score'))
        mlflow.log_metric("model id", 1)

        tracking_url_type_store = urlparse(mlflow.get_tracking_uri()).scheme

        # Registro del modelo en MLFlow
        if tracking_url_type_store != "file":
            mlflow.sklearn.log_model(model, "model", registered_model_name="Videogame_logistic_regression")
        else:
            mlflow.sklearn.log_model(model, "model")
        return experiment_model_log(model=model,name=run_name,classification_report=classification_report_model)

 obj=train_logistic_regression_model(x_train_small_transformed,x_test_small_transformed,y_train,y_test,run_name='Logistic Regression - Small')
experiment_model_list = [obj]

Test accuracy: 0.7662070200573066

 experiment_model_list.append(train_logistic_regression_model(x_train_medium_transformed,x_test_medium_transformed,y_train,y_test, max_iter=5000,run_name='Logistic Regression - Medium'))

Test accuracy: 0.8249462750716332

 experiment_model_list.append(train_logistic_regression_model(x_train_large_transformed,x_test_large_transformed,y_train,y_test, max_iter=5000,run_name='Logistic Regression - Large'))

Test accuracy: 0.8604047277936963

 def train_SVC_model(x_train, x_test, y_train, y_test, kernel='linear', run_name=''):
    with mlflow.start_run(run_name=run_name):
        #Definición de modelo con algoritmo
        model = SVC(kernel=kernel)

        #Entrenamiento del modelo
        model.fit(x_train, y_train)

        #Obtenemos predicciones
        #train_predict = model.predict(x_train)
        test_predict = model.predict(x_test)

        #Obtenemos métricas
        print(f"Test accuracy: {accuracy_score(y_test, test_predict)}")
        classification_report_model=classification_report(y_test, test_predict,output_dict=True)

        #Registro de Parametros en MLFlow
        mlflow.log_param("kernel", kernel)

        #Registro de Metricas en MLFlow
        mlflow.log_metric("Accuracy", classification_report_model.get('accuracy'))
        mlflow.log_metric("Precision", classification_report_model.get('macro avg').get('precision'))
        mlflow.log_metric("Recall", classification_report_model.get('macro avg').get('recall'))
        mlflow.log_metric("f-1 score", classification_report_model.get('macro avg').get('f1-score'))
        mlflow.log_metric("model id", 2)

        tracking_url_type_store = urlparse(mlflow.get_tracking_uri()).scheme

        # Registro del modelo en MLFlow
        if tracking_url_type_store != "file":
            mlflow.sklearn.log_model(model, "model", registered_model_name="Videogame_SVC")
        else:
            mlflow.sklearn.log_model(model, "model")
        return experiment_model_log(model=model,name=run_name,classification_report=classification_report_model)

 experiment_model_list.append(train_SVC_model(x_train_small_transformed,x_test_small_transformed,y_train,y_test,run_name='SVC - Small'))

Test accuracy: 0.7635207736389685

 experiment_model_list.append(train_SVC_model(x_train_medium_transformed,x_test_medium_transformed,y_train,y_test,run_name='SVC - Medium'))

Test accuracy: 0.825304441260745

 experiment_model_list.append(train_SVC_model(x_train_large_transformed,x_test_large_transformed,y_train,y_test,run_name='SVC - Large'))

Test accuracy: 0.8660458452722063

 def train_random_forest_model(x_train, x_test, y_train, y_test,n_estimators=100, max_depth=None, run_name=''):
    with mlflow.start_run(run_name=run_name):
        #Definición de modelo con algoritmo
        model = RandomForestClassifier(n_estimators=n_estimators,max_depth=max_depth)

        #Entrenamiento del modelo
        model.fit(x_train, y_train)

        #Obtenemos predicciones
        #train_predict = model.predict(x_train)
        test_predict = model.predict(x_test)

        #Obtenemos métricas
        print(f"Test accuracy: {accuracy_score(y_test, test_predict)}")
        classification_report_model=classification_report(y_test, test_predict,output_dict=True)

        #Registro de Parametros en MLFlow
        mlflow.log_param("n_estimators", n_estimators)
        mlflow.log_param("max_depth", max_depth)

        #Registro de Metricas en MLFlow
        mlflow.log_metric("Accuracy", classification_report_model.get('accuracy'))
        mlflow.log_metric("Precision", classification_report_model.get('macro avg').get('precision'))
        mlflow.log_metric("Recall", classification_report_model.get('macro avg').get('recall'))
        mlflow.log_metric("f-1 score", classification_report_model.get('macro avg').get('f1-score'))
        mlflow.log_metric("model id", 3)

        tracking_url_type_store = urlparse(mlflow.get_tracking_uri()).scheme

        # Registro del modelo en MLFlow
        if tracking_url_type_store != "file":
            mlflow.sklearn.log_model(model, "model", registered_model_name="Videogame_random_forest")
        else:
            mlflow.sklearn.log_model(model, "model")
        return experiment_model_log(model=model,name=run_name,classification_report=classification_report_model)

 experiment_model_list.append(train_random_forest_model(x_train_small_transformed,x_test_small_transformed,y_train,y_test,n_estimators=500,run_name='Random Forest - Small'))

Test accuracy: 0.8250358166189111

 experiment_model_list.append(train_random_forest_model(x_train_medium_transformed,x_test_medium_transformed,y_train,y_test,n_estimators=500,run_name='Random Forest - Medium'))

Test accuracy: 0.8504656160458453

 experiment_model_list.append(train_random_forest_model(x_train_large_transformed,x_test_large_transformed,y_train,y_test,n_estimators=500,run_name='Random Forest - Large'))

Test accuracy: 0.8640759312320917

 # Obtención de registro de pruebas
runs = mlflow.search_runs(experiment_ids=0)
runs_clean=runs

#Columnas que contiene dataframe
#print(runs_clean.columns)

#se filtran registros que se hayan terminado
runs_clean = runs_clean[runs_clean.iloc[:,2] == "FINISHED"]

# se filtran solo las columnas que interesa visualizar
runs_clean = runs_clean.iloc[:,[0,6,7,8,9,10,19]]

# Se obtienen dimensiones de dataframe
#print(runs_clean.shape)

#Mostramos resultados
runs_clean.head(10)

 #se obtiene indice de registro con mayor f1 score
index_f1score_max = np.argmax(runs_clean.iloc[:,1])

#se obtiene indice de registro con mayor accuracy
index_acc_max = np.argmax(runs_clean.iloc[:,2])

#se obtiene indice de registro con mayor recall
index_recall_max = np.argmax(runs_clean.iloc[:,3])

#se obtiene indice de registro con mayor precision
index_precision_max = np.argmax(runs_clean.iloc[:,4])

#se obtiene los registros con mejor desempeño para cada métrica y su model correspondiente
print('F1 Score Max - Run ID:',runs_clean.iloc[index_f1score_max,0],', Model: ',runs_clean.iloc[index_f1score_max,6],', Value:',runs_clean.iloc[index_f1score_max,2])
print('Accuracy Max - Run ID:',runs_clean.iloc[index_acc_max,0],', Model:',runs_clean.iloc[index_acc_max,6],', Value:',runs_clean.iloc[index_acc_max,2])
print('Recall Max - Run ID:',runs_clean.iloc[index_recall_max,0],', Model:',runs_clean.iloc[index_recall_max,6],', Value:',runs_clean.iloc[index_recall_max,2])
print('Precision Max - Run ID:',runs_clean.iloc[index_precision_max,0],', Model:',runs_clean.iloc[index_precision_max,6],', Value:',runs_clean.iloc[index_precision_max,2])

F1 Score Max - Run ID: 5ba79e719c524918b3febf002ec1452c , Model:  SVC - Large , Value: 0.865942554025767
Accuracy Max - Run ID: 5ba79e719c524918b3febf002ec1452c , Model: SVC - Large , Value: 0.865942554025767
Recall Max - Run ID: 5ba79e719c524918b3febf002ec1452c , Model: SVC - Large , Value: 0.865942554025767
Precision Max - Run ID: 5ba79e719c524918b3febf002ec1452c , Model: SVC - Large , Value: 0.865942554025767

 def metrics_bar_graph(experiments=None):
    key_macro = 'macro avg'
    metricas=['Accuracy', 'Precision', 'Recall', 'f1-Score']
    fig_data = dict({'data':[{}]})

    #Iteracion en cada uno de los experimentos
    for experimento in experiments:
        data = dict({
            'type': 'bar',
            'name': str(experimento.name), 
            'x': metricas,
            'y': [  experimento.classification_report.get(metricas[0].lower()), 
                    experimento.classification_report.get(key_macro).get(metricas[1].lower()), 
                    experimento.classification_report.get(key_macro).get(metricas[2].lower()),
                    experimento.classification_report.get(key_macro).get(metricas[3].lower())]})
        fig_data['data'].append(data)

    fig = go.Figure(data=fig_data)
    # Change the bar mode
    fig.update_layout(barmode='group')
    fig.update_traces(texttemplate='%{y:.3}')
    fig.update_yaxes(title_text="Resultados", range=[0,1], titlefont=dict(size=30)) #tickvals=[0, 0.2, 0.725, 0.750, 0.8,1] 
    fig.update_xaxes(title_text="Metricas", titlefont=dict(size=30))

    fig.show(renderer='notebook')

 def metrics_line_graph(vectorize='Small' ,experiments=None):
    metricas=['Accuracy', 'Precision', 'Recall', 'f1-Score']
    key_macro = 'macro avg'
    fig_data = dict({'data':[{}]})
    vec1, vec2, vec3 = 0,3,5

    if(vectorize.lower() == 'small'):
        vec1, vec2, vec3 = 0,3,6
    elif(vectorize.lower() == 'medium'):
        vec1, vec2, vec3 = 1,4,7
    elif(vectorize.lower() == 'large'):
        vec1, vec2, vec3 = 2,5,8

    data = dict({
        'name': metricas[0], 
        'x': [  str(experiments[vec1].name),
                str(experiments[vec2].name),
                str(experiments[vec3].name)],
        'y': [  experiments[vec1].classification_report.get(metricas[0].lower()), 
                experiments[vec2].classification_report.get(metricas[0].lower()),
                experiments[vec3].classification_report.get(metricas[0].lower())]})
    fig_data['data'].append(data)

    for i in range(1, 4):
        data = dict({
            'name': metricas[i], 
            'x': [  str(experiments[vec1].name),
                    str(experiments[vec2].name),
                    str(experiments[vec3].name)],
            'y': [  experiments[vec1].classification_report.get(key_macro).get(metricas[i].lower()), 
                    experiments[vec2].classification_report.get(key_macro).get(metricas[i].lower()),
                    experiments[vec3].classification_report.get(key_macro).get(metricas[i].lower())]})
        fig_data['data'].append(data)
    
    fig = go.Figure(data=fig_data)
    # Change the bar mode
    fig.update_layout(barmode='group', title=vectorize + ' Vectorizer')
    fig.update_traces(texttemplate='%{y}')
    fig.update_yaxes(title_text="Resultados", range=[0,1], titlefont=dict(size=30), autorange=True)
    fig.update_xaxes(titlefont=dict(size=30), autorange=True)

    fig.show(renderer='notebook')

 metrics_bar_graph(experiment_model_list)

 metrics_line_graph(vectorize='Small', experiments=experiment_model_list)
metrics_line_graph(vectorize='Medium', experiments=experiment_model_list)
metrics_line_graph(vectorize='Large', experiments=experiment_model_list)

Nombre Columna	Descripcion	Utilizar
reviewerID	ID of the reviewer, e.g. A2SUAM1J3GNN3B	No
asin	ID of the product, e.g. 0000013714	No
reviewerName	name of the reviewer	No
vote	helpful votes of the review	No
style	a disctionary of the product metadata, e.g., "Format" is "Hardcover"	No
reviewText	text of the review	Si
overall	rating of the product	Si
summary	summary of the review	No
unixReviewTime	time of the review (unix time)	No
reviewTime	time of the review (raw)	No
image	images that users post after they have received the product	No

	reviewerID	asin	reviewerName	helpful	reviewText	overall	summary	unixReviewTime	reviewTime
0	A2HD75EMZR8QLN	0700099867	123	[8, 12]	Installing the game was a struggle (because of...	1.0	Pay to unlock content? I don't think so.	1341792000	07 9, 2012
1	A3UR8NLLY1ZHCX	0700099867	Alejandro Henao "Electronic Junky"	[0, 0]	If you like rally cars get this game you will ...	4.0	Good rally game	1372550400	06 30, 2013
2	A1INA0F5CWW3J4	0700099867	Amazon Shopper "Mr.Repsol"	[0, 0]	1st shipment received a book instead of the ga...	1.0	Wrong key	1403913600	06 28, 2014
3	A1DLMTOTHQ4AST	0700099867	ampgreen	[7, 10]	I got this version instead of the PS3 version,...	3.0	awesome game, if it did not crash frequently !!	1315958400	09 14, 2011
4	A361M14PU2GUEG	0700099867	Angry Ryan "Ryan A. Forrest"	[2, 2]	I had Dirt 2 on Xbox 360 and it was an okay ga...	4.0	DIRT 3	1308009600	06 14, 2011
5	A2UTRVO4FDCBH6	0700099867	A.R.G.	[0, 0]	Overall this is a well done racing game, with ...	4.0	Good racing game, terrible Windows Live Requir...	1368230400	05 11, 2013
6	AN3YYDZAS3O1Y	0700099867	Bob	[11, 13]	Loved playing Dirt 2 and I thought the graphic...	5.0	A step up from Dirt 2 and that is terrific!	1313280000	08 14, 2011
7	AQTC623NCESZW	0700099867	Chesty Puller	[1, 4]	I can't tell you what a piece of dog**** this ...	1.0	Crash 3 is correct name AKA Microsoft	1353715200	11 24, 2012
8	A1QJJU33VNC4S7	0700099867	D@rkFX	[0, 1]	I initially gave this one star because it was ...	4.0	A great game ruined by Microsoft's account man...	1352851200	11 14, 2012
9	A2JLT2WY0F2HVI	0700099867	D. Sweetapple	[1, 1]	I still haven't figured this one out. Did ever...	2.0	Couldn't get this one to work	1391817600	02 8, 2014

	review_text	sentiment
0	Installing the game was a struggle (because of...	1
1	If you like rally cars get this game you will ...	4
2	1st shipment received a book instead of the ga...	1
3	I got this version instead of the PS3 version,...	3
4	I had Dirt 2 on Xbox 360 and it was an okay ga...	4

	run_id	metrics.Precision	metrics.f-1 score	metrics.Accuracy	metrics.Recall	metrics.model id	tags.mlflow.runName
0	5c2905a2d5ce49619e6d583e4f3e64c4	0.864022	0.864010	0.864076	0.863999	3.0	Random Forest - Large
1	fdd1c35d0ffa4f9aa77449cfa9b9981e	0.850423	0.850383	0.850466	0.850352	3.0	Random Forest - Medium
2	b3816a54eb524c79baf7398e02c4e0ca	0.824949	0.824969	0.825036	0.824995	3.0	Random Forest - Small
3	5ba79e719c524918b3febf002ec1452c	0.866110	0.865943	0.866046	0.865848	2.0	SVC - Large
4	47ad4eca8d3646228c7818015a4f1fdb	0.825268	0.825277	0.825304	0.825410	2.0	SVC - Medium
5	e86d0e0e077f499883d41797b80f622f	0.763906	0.763096	0.763521	0.762965	2.0	SVC - Small
6	b432ae9de16a4bdc940f9f41e2955ad4	0.860564	0.860271	0.860405	0.860137	1.0	Logistic Regression - Large
7	7b9419f4a67e490fa4b43b9db2473a00	0.825141	0.824749	0.824946	0.824605	1.0	Logistic Regression - Medium
8	f694101147b144cb859bcef5bcfb2654	0.766516	0.765825	0.766207	0.765694	1.0	Logistic Regression - Small
10	bebd67430e594db2ac99af3b8e8da148	0.864862	0.864575	0.864703	0.864441	1.0	Logistic Regression - Large

Marcelino Mayorga Quesada ¶

Objetivo¶

Estrategia y Resumén Técnico¶

Arquitectura de la solución¶

Importación de Librerías¶

Fuente de datos¶

Información¶

Definición de Fuente de Datos: Reviews de Video Games¶

Consideraciones¶

Carga de datos¶

Análisis exploratorio de los datos¶

Descripción general del dataset¶

Generación de Reportes y Análisis Univariantes y Multivariantes¶

Observaciones y Análisis¶

Procesado de datos¶

Limpiado de datos¶

Generación de nuevo dataset de trabajo¶

Eliminación de muestras de sentiment = 3¶

Balanceo de datos¶

Categorización de reviews¶

Preparación de los conjuntos de train y test¶

Normalización y Extracción de características.¶

Entrenamiento de un modelo de análisis de sentimiento¶

Logistic Regression¶

C-Support Vector Classification(SVC)¶

RandomForestClassifier¶

Validación y listado de métricas registradas en MLFLOW¶

Listado de métricas¶

Evaluación de modelos¶

Visualización de métricas¶

Métodos para graficar resultados¶

Gráfico de Barras - Resultados de Métricas por Modelo¶

Gráfico de Líneas - Resultados de Métricas por Vectorizer y Modelo de Entrenamiento¶

	review_text	sentiment	sentiment_label
138127	This game looks like a driving game, but it is...	1	neg
169555	It's a cool idea of playing 4 games from previ...	1	neg
109323	I will never buy a game with invasive, crippli...	1	neg
64521	The item pictured on Amazon is not the item yo...	1	neg
187417	So apparently I wasted my money on this game, ...	1	neg

Marcelino Mayorga Quesada¶

Objetivo¶

Estrategia y Resumén Técnico¶

Arquitectura de la solución¶

Importación de Librerías¶

Fuente de datos¶

Información¶

Definición de Fuente de Datos: Reviews de Video Games¶

Consideraciones¶

Carga de datos¶

Análisis exploratorio de los datos¶

Descripción general del dataset¶

Generación de Reportes y Análisis Univariantes y Multivariantes¶

Observaciones y Análisis¶

Procesado de datos¶

Limpiado de datos¶

Generación de nuevo dataset de trabajo¶

Eliminación de muestras de sentiment = 3¶

Balanceo de datos¶

Categorización de reviews¶

Preparación de los conjuntos de train y test¶

Normalización y Extracción de características.¶

Entrenamiento de un modelo de análisis de sentimiento¶

Logistic Regression¶

C-Support Vector Classification(SVC)¶

RandomForestClassifier¶

Validación y listado de métricas registradas en MLFLOW¶

Listado de métricas¶

Evaluación de modelos¶

Visualización de métricas¶

Métodos para graficar resultados¶

Gráfico de Barras - Resultados de Métricas por Modelo¶

Gráfico de Líneas - Resultados de Métricas por Vectorizer y Modelo de Entrenamiento¶

Marcelino Mayorga Quesada ¶