Introduction

The evaluation is about the sentimen analysis over the review and stars rating restaurant in Last Vegas city.. The YELP dataset is very resouceful which provides the valuable criteria over 61,184 unique records for business , 1,569,264 records for reviews and 495,107 records for the tips. 2 tables been discarded for now which is user details and the checkin information.The GPS longitute and latitude available inside the business dataset which provides very useful information about its geolocation. The stars value gives the feedback from the customer which might be positive , negative or neutral. The findings offer exemplary big data analysis methods as the evaluation of socially mediated urban space associated with the pattern classification of textual information inside the reviews and tips in relation with business dataset.

From the total sample inside business dataset , Las Vegas City is the top 5 location with the most review counted as follows :-

Top 5 City Reviews and Categories
business_categories city review_count
[Breakfast & Brunch, Steakhouses, French, Restaurants] Las Vegas 4578
[Sandwiches, Restaurants] Las Vegas 3984
[Buffets, Restaurants] Las Vegas 3828
[Buffets, Restaurants] Las Vegas 3046
[American (Traditional), Restaurants] Las Vegas 3007
[Buffets, Restaurants] Las Vegas 2949

The similarities between the business types is the keyword Restaurant

Methods and Data

The dataset is obtained from the YELP website (http://www.yelp.com/dataset_challenge) and extracted. The format for the dataset is in JSON . JSON need special techniques to parse and read from it. Apache Hive is the best component which is capable read this format . Since the dataset required to have a good machine in term of CPU and memory , we push this dataset to work inside Hadoop which Map-Reduce can be used as the framework for the filtering and cleaning over large size of the dataset. Hive is compatible to use scripting parameter similar to SQL and this is very suitable for speed up the entire development work. Hive also support for the complex data type and STRUCT is used to handle the JSON complex type for the table creation inside Hive .

For the basic analysis , this evaluation requires a fair amount time to know about the dataset abd performaing exploratory analysis. But, now we only focus on the textual information which mostly inside the review and tips dataset in conjuction with the business and user information. This will tackle some of the questions such as :-

Below is the basic summary for the questions above :-

from the above information we can make assumption that the result shows the common food that always mention by the reviewer are chicken , sushi , burger , pizza , cheese , salad , rice and sauce. We use this list as the base of common food can be relate with the emotion of the reviewer. List of emotion that can be identified such as friendly , best , well , and rice . So, we interested with the food type listed and we want to have some idea what are their possible relationship . To achive this , is to use the word-graph techniques and the result as follows :-

The average no of words inside the review and tips messages is xxxxxx . the average word length is xxxx. emotion type that might contain inside the review and tips messages is like xxxxx , What is the most frequent words or terms inside it is xxxxxxxxxxxx

Other interesting findings in this evaluation is to classify the reviewers ratings and the tips provided. The idea is to calculate the sentimen score for each messages so we can know how positive and negative the messages. Below is the formula of the how to calculate the score :-

      Score  =  Number of positive words  -  Number of negative words`

The lexicon is in English and the reference for the positive and negative words is reference from (https://github.com/SamPortnow/Depression_Prevention_Program/tree/master/bato/assets).

  1. Method Used - Classification , Bayes

Results

Summary of Las Vegas City Restaurant No. Of Review
Min. 1st Qu. Median Mean 3rd Qu. Max.
3 9 26 96 87 4578

To reduce the size of the sample , average size of numbers of message is the minimal size which is around 390. And the numbers of group identified around 1000

From the summary show that the Median is 26 and we choose 26 as the minimal sample for this evaluation. The median better than mean because of it is a symmetrical statistic and more resistant to errors.

xxx

Figure 1 : Venue with greater than 390 review messages

Bagitau list of restaurant yang banyak impact dan types of food yang ada sesama mereka kalau ada relationship

The population is because the location is very strategic and nearest to the airport

Figure 2 : Venue with greater than 390 review messages

Positive correlated with the positive emotion words

Figure 3 : Venue with weighted by the numbers of keywords

Sentiment analysis on keyword left 70% of the dataset with neutral sentiment. Sentiment was found to be 74% positive in nature which corresponds to about 22% of the total sample In contrast,26% was negative in nature which corresponds to only 8% of the total sample

Figure 4 : Venue with weighted by the numbers of keywords , with positive weight Figure 5 : Venue with weighted by the numbers of keywords , with negative weight Figure 6 : Total sentiment classification with positive , negative and neutral Figure 7 : Cloudwords postive and negative

Figure 8 : Duration - Negative vs Positive + Neutral Figure 9 : Comparison by month

Discussion

The issues raise a significant question *

The HEAD records for business types , average ratings and the average review count as follows :

Las Vegas City Restaurant
name stars review_count
Mon Ami Gabi 4 4578
Earl of Sandwich 4.5 3984
Wicked Spoon 3.5 3828
Bacchanal Buffet 4 3046
Serendipity 3 3 3007
The Buffet 3.5 2949

Is there any relation between the business type ?

The HEAD records for business types , average ratings and the average review count as follows :

Tips

Cerita techniques yang dipakai di sini

https://sites.google.com/site/miningtwitter/questions/talking-about/given-topic