🔥 A review sentiment prediction analysis for any dataset using h2o
This application shows the predicted value of your review comment to be positive or negative based on the dataset at hand and their previous classifications. Please run as one complete block to include UI elements and the underlying text mining methods. Please see Review App.py for undelying h2o logic. For example purposes, you can use this UCI Machine Learning Datasets] which has the following datasets
| Datasets |
|---|
| Amazon Reviews |
| Yelp Restaurant Reviews |
| IMDB Movies reviews |
The code block contains both the ui elements and the underlying classification algorithm In this case : h2o.ai. Please note that the prediction depends purely on the supervised set input to the application
#using tkinter for UI
from tkinter import *
from tkinter import filedialog
#inititate the input file path and output
tsv_file=''
output_value=0.0
def get_pred():
jobname=entry5.get()
return jobname
#File :ask to open
def UploadAction(event=None):
filename = filedialog.askopenfilename()
tsvfile=filename.get()
print('Selected:', filename)
def AssignAction(event=None):
filename = filedialog.askopenfilename()
#tsvfile=filename
return filename
#prediction algorithm
def predict(job_title,w2v, gbm):
import h2o
words = tokenize(h2o.H2OFrame(job_title).ascharacter())
job_title_vec = w2v.transform(words, aggregate_method="AVERAGE")
#job_title_vec_2= job_title.cbind(job_title_vec)
print(gbm.predict(job_title_vec))
if gbm.predict(job_title_vec)>0.5:
print("Most likely Negative")
else:
print("Probably positive")
def submitted():
predx=get_pred()
op=(predict([predx], w2v_model, gbm_model))
print(predict([predx], w2v_model, gbm_model))
# tokenize to split each word of review
def tokenize(sentences, stop_word = ["ax","i","you","edu","s","t","m","subject","can","lines","re","what","there","all","we","one","the","a","an","of","or","in","for","by","on","but","is","in","a","not","with","as","was","if","they","are","this","and","it","have","from","at","my","be","by","not","that","to","from","com","org","like","likes","so"]
):
tokenized = sentences.tokenize("\\W+")
tokenized_lower = tokenized.tolower()
tokenized_filtered = tokenized_lower[(tokenized_lower.nchar() >= 2) | (tokenized_lower.isna()),:]
tokenized_words = tokenized_filtered[tokenized_filtered.grep("[0-9]",invert=True,output_logical=True),:]
tokenized_words = tokenized_words[(tokenized_words.isna()) | (~ tokenized_words.isin(["ax","i","you","edu","s","t","m","subject","can","lines","re","what","there","all","we","one","the","a","an","of","or","in","for","by","on","but","is","in","a","not","with","as","was","if","they","are","this","and","it","have","from","at","my","be","by","not","that","to","from","com","org","like","likes","so"]
)),:]
return tokenized_words
# The application interface
def analyze():
tsv_file=AssignAction()
import time
time.sleep(10)
import h2o
h2o.init(max_mem_size="2G")
label3 = Label(root, text="Building Model")
label3.pack()
import pandas as pd
import numpy as np
csv_table=pd.read_table(tsv_file,sep='\t',header=None,names = ["Review", "Response"] )
csv_table.to_csv('reviews.csv',index=False)
#print(type(csv_table))
csv_table
reviews = h2o.h2o.H2OFrame(csv_table)
words = tokenize(reviews["Review"])
from h2o.estimators.word2vec import H2OWord2vecEstimator
w2v_model = H2OWord2vecEstimator(vec_size = 100, model_id = "w2v.hex")
w2v_model.train(training_frame=words)
review_vecs = w2v_model.transform(words, aggregate_method = "AVERAGE")
#ext_reviews
ext_reviews = reviews.cbind(review_vecs)
data_split = ext_reviews.split_frame(ratios=[0.8])
ext_train = data_split[0]
ext_test = data_split[1]
#print("Build a basic GBM model")
gbm_model = h2o.estimators.gbm.H2OGradientBoostingEstimator()
gbm_model.train(x = ext_reviews.names,
y="Response",
training_frame = data_split[0],
validation_frame = data_split[1])
#output_value=predict([entry5.get()], w2v_model, gbm_model)
#label7 = Label(root, text=str(output_value))
#label7.pack()
x=entry5.get()
y=h2o.H2OFrame([x])
y.col_names = ['Review']
return(predict([y], w2v_model, gbm_model))
#root.mainloop()
root = Tk()
root.title("h2o Data Analytics")
#Label 1
label1 = Label(root,text = 'Prediction of Comments')
label1.pack()
label1.config(justify = CENTER)
label5 = Label(root, text="Enter a comment that you want to predict")
label5.pack()
#label1.config(justify = CENTER)
entry5 = Entry(root, width = 100)
entry5.pack()
label4 = Label(root, text="Upload review data file to start the analysis based on the data")
label4.pack()
button6 = Button(root, text = 'Start Now')
button6.pack()
button6.config(command = analyze)
root.mainloop()Checking whether there is an H2O instance running at http://localhost:54321. connected.
| H2O cluster uptime: | 4 hours 52 mins |
| H2O cluster timezone: | America/New_York |
| H2O data parsing timezone: | UTC |
| H2O cluster version: | 3.22.0.1 |
| H2O cluster version age: | 15 days |
| H2O cluster name: | H2O_from_python_narensham_b4de52 |
| H2O cluster total nodes: | 1 |
| H2O cluster free memory: | 1.472 Gb |
| H2O cluster total cores: | 8 |
| H2O cluster allowed cores: | 8 |
| H2O cluster status: | locked, healthy |
| H2O connection url: | http://localhost:54321 |
| H2O connection proxy: | None |
| H2O internal security: | False |
| H2O API Extensions: | XGBoost, Algos, AutoML, Core V3, Core V4 |
| Python version: | 3.6.4 final |
Parse progress: |█████████████████████████████████████████████████████████| 100%
word2vec Model Build progress: |██████████████████████████████████████████| 100%
gbm Model Build progress: |███████████████████████████████████████████████| 100%
Parse progress: |█████████████████████████████████████████████████████████| 100%
| Review |
|---|
| What is this about, I have no clue |
Parse progress: |█████████████████████████████████████████████████████████| 100%
gbm prediction progress: |████████████████████████████████████████████████| 100%
| predict |
|---|
| 0.338029 |
gbm prediction progress: |████████████████████████████████████████████████| 100%
Most likely Negative
The above statement specifies the likely classification based on the dataset provided.