This was a project of a data mining class I completed in college.

Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.

README.md -- You're reading it now. Makefile -- Contains various recipies for creating directories, zipping the project together, and running large batches of the KNN algorithm.

The usage message for any of the command line programs can be found by using the '-h' flag. For example: 'python3 runKNN.py -h'.

createProfDict.py -- Creates the profDicts.pkl data structure. createProfVectors.py -- Creates professor token count vectors. findCorrelations.py -- Finds pearson correlation between token pair count ratios and rating value. Outputs a plot. findPids.py -- Creates pids files, which hold the pids of professors with single reviews, and small reviews. getStatistics.py -- Finds some basic information about the distrobution of reviews and their sizes among professors. This also calculates the mean absolute error of the naive prediction approach, which is always predicting the mean rating. runFFNN.py -- Trains a feedforward neural network on professor vectors. Excludes single and small vectors to use as validation set. runKNN.py -- Runs the KNN algorithm to predict rating for all professor vectors.

Counting.py -- Contains functions used in countTokens.py FfnnAlgorithm.py -- Contains two FFNN topologies. FileNames.py -- Contains constants and functions to hold and create filenames. KnnAlgorithm.py -- Contains functions to run KNN and create the similarity matrix. PlotData.py -- Contains functions to create plots to visualize data. ReadFromFile.py -- Reads various data structures from file. Stats.py -- Contains functions used in getStatistics.py VectorProcessing.py -- Contains functions concerned with vector creation and modification. WriteToFile.py -- Writes various data structures to file.

data/polyratings.sql -- SQL Dump of the datasets used. data/profDicts.pkl -- Holds a list of profDicts, as described in the report. data/professors.pkl -- Holds the professor table of the polyratings dataset as a list of dictionaries. data/reviews.pkl -- Holds the review table of the polyratings dataset as a list of dictionaries. data/stopwords-mysql.txt -- Contains stopwords used during initial vector processing.

data/correlations/ -- Purposely left empty to reduce the size of the zip file. For a plot of these correlations refer to the report. data/knnPredictions/ -- Directory included in zip, though contains no data. This directory holds predictions of professor vectors for various k values and will be populated with runs of runKNN.py. Predictions were excluded from the zip file to reduce the files size. data/pids/ -- Contans pids of professors with only one review and with small aggrigate reviews, as defined in the report. data/similarityMatricies/ -- Contains similarity matricies for different vectors. data/tokenCounts/ -- Contains token count files for various vector configurations as well as profTokenDict pickled datastructures, which are an intermediary datastructure for vector creation. data/vectors/ -- Contians professor vector files.

The directory structure for figures is included in the zip file, though no figures are included. This decision was made in order to reduce the size of the zip file. Instructions to create figures are below.

figures/counts.png -- Plots the distrobution of review size, number of reviews per professor, and aggrigate review size for each professor. To create this file, run 'python3 getStatistics.py -save'.

figures/correlations/ -- Normally contains correlation plots. These are also found in the report. To create the run findCorrelations.py with the '-save' flag. Note: If the correlations file hasn't been created beforehand, this will take a very long time to run, perhaps up to six hours. figures/knnAccuracy/ -- Contains plots on the accuracy of a KNN run for all professor vectors, professor vectors with one review, and small professor vectors. To create these plots, run runKNN.py with the '-save' flag. figures/tokenCount/ -- Contains plots showing the distrobution of tokens for different token processing methods. To create this plot run countTokens.py with the '-save' flag.