avinashkalivarapu / restaurantrecommendationsystem Goto Github PK

Classifiers
● Linear SVM ​
: We trained and tested our dataset using Linear SVM classifier.
Linear SVM is the newest extremely fast machine learning (data mining) algorithm
for solving multi class classification problems from ultra large data sets that implements
an original proprietary version of a cutting plane algorithm for designing a linear support
vector machine. Linear SVM is a linearly scalable routine meaning that it creates an SVM
model in a CPU time which scales linearly with the size of the training data set.
● SVM with RBF kernel : ​
We trained and tested our dataset using SVM with
RBF kernel classifier.
RBF network can be used find a set weights for a curve fitting problem. The
weights are in higher dimensional space than the original data. I Learning is equivalent to
finding a surface in high dimensional space that provides the best fit to training data. I
Hidden layers provide a set of functions that constitute an arbitrary basis for input
patterns when they are expanded to the hidden space; these functions are called radial
basis functions.
● Logistic Regression : ​
We trained and tested our dataset using logistic
regression.
Logistic Regression is a special type of regression where binary response variable
is related to a set of explanatory variables, which can be discrete and/or continuous. The
important point here to note is that in linear regression, the expected values of the
response variable are modeled based on combination of values taken by the predictors.
In logistic regression Probability or Odds of the response taking a particular value is
modeled based on combination of values taken by the predictors.
6● Random Forest : ​
We trained and tested our dataset using random forest.
Random forests is a notion of the general technique of random decision
[2] ​
forests [1]
​ ​
that are an ensemble learning method for classification,regression and other
tasks, that operate by constructing a multitude of ​
decision trees ​
at training time and
outputting the class that is the ​
mode ​
of the classes (classification) or mean prediction
(regression) of the individual trees. Random decision forests correct for decision trees'
habit of overfitting to their training set.
● We also built a new classifier method using ​
Neural Networks ​
for calculating the
weights for the features .
○ A net weight score is calculated for each of the available restaurants in the
dataset.
○ Net Weight score for every restaurant is calculated by taking all the
features of that user and restaurant that are mentioned above
○ Weights are given to each and every taken feature.
○ A weight for every feature is found by backpropagation of the training data
( data of last few months) by building a neural network.
○ Given a user(U) and a restaurant(R), one should find out whether the user
likes it or not.
○ All the restaurants for which that particular user has given reviews are
considered and an average net weight score is calculated.
○ This calculated average net score is taken as a threshold(t).
○ The net weight score(w) for this restaurant(R) will be known as the weight
scores for all the restaurants are calculated in the beginning.
○ If we find w to be greater than the taken threshold t , we return a ‘yes’
7saying the user will like this restaurant.
○ If we find w to be lesser than the taken threshold t , we return a ‘no’ saying
the user will not like this restaurant.


Dataset
The data that we used in this project was obtained from the Yelp Dataset challenge. The
dataset contains five different tables: User, Business, Review, Check-In and Tips. The data
has 27257 restaurants, 552339 users, 55569 check-Ins, 591864 tips and 2225213 reviews.From this Yelp dataset, we took the latest 1 month data as test dataset. Apart from the
test dataset, last 3 months data was taken as training dataset. Remaining part of the data
apart from the test and training is used for calculating derived features(historical data).


FURTHER DETAILS ARE SPECIFIED IN THE REPORT