Merge pull request 'add oneR to main' (#2) from oneR into main

Reviewed-on: #2

.gitignore

@@ -34,3 +34,6 @@
 
 # Build directory
 build
+
+# Clang files
+.cache

README.md

@@ -1,5 +1,12 @@
 # arff-mining
 
+## Author, Date, Description
+- Gregory Crawford
+- Date: 2024-04-17
+- Description: Read data from ARFF file and generate OneR output 
+- Performance summary at the bottom of this README
+
+
 ## Compiling the project
 
 Prerequisites
@@ -10,6 +17,7 @@ In order to compile the project, simply run these two commands:
     cmake -B build -S .
     cmake --build build
 
+
 ## Running the Project
 
 The programs should now be compiled at ./build/bin/
@@ -18,3 +26,52 @@ ARFF:
 ```plain
 build/bin/arff
 ```
+
+One Rule:
+```plain
+build/bin/onerule
+```
+
+
+## Performance Summary
+
+***contactLenses***
+```
+	tear-prod-rate:
+		normal ---> soft
+		reduced ---> none
+Error rate: 7/24
+build/bin/onerule  0.00s user 0.00s system 85% cpu 0.002 total
+```
+
+***restaurants***
+```
+	Pat:
+		Full ---> No
+		None ---> No
+		Some ---> Yes
+Error rate: 2/12
+build/bin/onerule  0.00s user 0.00s system 85% cpu 0.001 total
+```
+
+***soybean***
+```
+	fruit-spots:
+		absent ---> alternarialeaf-spot
+		brown-w/blk-specks ---> anthracnose
+		colored ---> frog-eye-leaf-spot
+		distort ---> alternarialeaf-spot
+		dna ---> brown-stem-rot
+Error rate: 385/683
+build/bin/onerule  0.02s user 0.00s system 98% cpu 0.024 total
+```
+
+***weather***
+```
+	outlook:
+		overcast ---> yes
+		rainy ---> yes
+		sunny ---> no
+Error rate: 4/14
+build/bin/onerule  0.00s user 0.00s system 87% cpu 0.001 total
+```

src/CMakeLists.txt

@@ -1 +1,2 @@
 add_subdirectory(arff)
+add_subdirectory(onerule)

src/arff/CMakeLists.txt

@@ -1,5 +1,6 @@
 add_executable(arff
-    ./main.cpp
+    main.cpp
-    ./arff.cpp
+    arff.cpp
+    log.cpp
 )
 

src/arff/arff.cpp

@@ -1,18 +1,17 @@
 #include "arff.hpp"
+#include "log.hpp"
 #include <iostream>
 #include <fstream>
 #include <sstream>
 #include <algorithm>
 
 namespace ARFF {
-    bool isVerbose = false;
-
     void ParseArguments(int argc, char* argv[]) {
         std::string argument_string;
         for (int i = 0; i < argc; ++i) {
             argument_string.assign(argv[i]);
             if (argument_string == "-v" || argument_string == "--verbose") { 
-                isVerbose = true;
+                debug::verbose = true;
             }
         }
     }
@@ -22,25 +21,13 @@ namespace ARFF {
         std::cout << "Please enter name of the data file:\t";
         std::cin >> filename;
         if (filename.empty()) {
-            LogError("ARFF/Setup", "No data filename provided, exiting...");
+            debug::Log(kError, "No data filename provided, exiting...");
             exit(1);
         }
         std::cout << std::endl;
         return filename;
     }
 
-    void LogInfo(const std::string location, const std::string message) {
-        if (!isVerbose) { return; }
-        std::cout << '[' << location << " - INFO] ";
-        std::cout << message << std::endl;
-    }
-
-    void LogError(const std::string location, const std::string message) {
-        if (!isVerbose) { return; }
-        std::cerr << '[' << location << " - ERROR] ";
-        std::cerr << message << std::endl;
-    }
-
     AttributeType::AttributeType(std::string attribute) {
         this->attribute = attribute;
     }
@@ -54,11 +41,15 @@ namespace ARFF {
         this->values.resize(size);
     }
 
+    AttributeEvaluation::AttributeEvaluation(AttributeType *attribute) {
+        this->currentAttribute = attribute;
+    }
+
     // Read entire data file and parse it
     void Arff::Read(std::string filename) {
         std::ifstream dataFile(filename);
         if (!dataFile.is_open()) {
-            LogError("ARFF/Read", "Unable to open file with name `" 
+            debug::Log(kError, "Unable to open file with name `" 
                     + filename + ", exiting...");
             exit(1);
         }
@@ -81,11 +72,16 @@ namespace ARFF {
         TestIntegrity();
     }
 
-    void Arff::Print(void) {
+    // Print generic data information
+    // Number of attributes and size of database
+    void Arff::PrintOverview(void) {
         std::cout << attributeList.size() << " attributes\n";
         std::cout << database.size() << " examples\n";
         std::cout << std::endl;
+    }
 
+    // Print full data information
+    void Arff::PrintData(void) {
         std::cout << "Attribute (#): values\n";
         for (AttributeType type : attributeList) {
             std::cout << type.attribute << " (" << type.values.size() << "):";
@@ -107,6 +103,19 @@ namespace ARFF {
         std::cout << std::endl;
     }
 
+    // Print result of applying OneR
+    // TODO: Create function
+    void Arff::OneR(void) {
+        AttributeEvaluation bestAttribute = _OneR();
+        debug::Log(kNone, "***Best 1-rule***");
+        debug::Log(kNone, "\t" + bestAttribute.currentAttribute->attribute + ':');
+        for (auto it : bestAttribute.rules) {
+            debug::Log(kNone, "\t\t" + it.first + " ---> " + it.second);
+        }
+        debug::Log(kNone, "Error rate: " + std::to_string(bestAttribute.totalError) + "/" + std::to_string(database.size()));
+    }
+
+
     // Add the attribute to the list
     void Arff::AddAttribute(std::string line) {
         std::stringstream parser(line);
@@ -120,12 +129,12 @@ namespace ARFF {
         if (token == "@relation" || token == "@RELATION") {
             parser >> token;
             relation = token;
-            LogInfo("ARFF/Attribute", "Relation set: " + relation);
+            debug::Log(kLog, "Relation set: " + relation);
             return;
         }
         parser >> token;
         attributeList.emplace_back(token);
-        LogInfo("ARFF/Attribute", "Added attribute: " + token);
+        debug::Log(kLog, "Added attribute: " + token);
         while (std::getline(parser, token, ',')) {
             // Clean token from outside pieces
             token.erase(std::remove(token.begin(), token.end(), ' '), token.end());
@@ -136,7 +145,7 @@ namespace ARFF {
             token.erase(std::remove(token.begin(), token.end(), '\r'), token.end());
             token.erase(std::remove(token.begin(), token.end(), '\n'), token.end());
             attributeList.back().AddValue(token);
-            LogInfo("ARFF/Attribute", "Added value: " + token);
+            debug::Log(kLog, "Added value: " + token);
         }
         // Additional missing value case
         attributeList.back().AddValue("?");
@@ -149,14 +158,14 @@ namespace ARFF {
         int id = 0;
         if (!database.empty()) { id = database.back().id + 1; }
         database.emplace_back(id, attributeList.size());
-        LogInfo("ARFF/Data", "Added id: " + std::to_string(database.back().id));
+        debug::Log(kLog, "Added id: " + std::to_string(database.back().id));
         for (int i = 0; i < attributeList.size(); ++i) {
             std::getline(parser, token, ',');
             token.erase(std::remove(token.begin(), token.end(), ' '), token.end());
             token.erase(std::remove(token.begin(), token.end(), '\r'), token.end());
             token.erase(std::remove(token.begin(), token.end(), '\n'), token.end());
             database.back().values.at(i) = token;
-            LogInfo("ARFF/Data", "Added instance value: " + token);
+            debug::Log(kLog, "Added instance value: " + token);
         }
     }
 
@@ -164,25 +173,81 @@ namespace ARFF {
         for (Instance instance : database) {
             int successCheck = 0;
             for (int i = 0; i < attributeList.size(); ++i) {
-                LogInfo("ARFF/Integrity", "Instance value tested: '" 
+                debug::Log(kTrace, "Instance value tested: '" 
                         + instance.values.at(i) + "'");
                 for (std::string value : attributeList.at(i).values) {
-                    LogInfo("ARFF/Integrity", "attributeList value: '" 
+                    debug::Log(kTrace, "attributeList value: '" 
                             + value + "'");
                     if (instance.values.at(i) == value) {
-                        LogInfo("ARFF/Integrity", "Value found: " + value);
+                        debug::Log(kTrace, "Value found: " + value);
                         ++successCheck;
                         break;
                     }
                 }
             }
             if (successCheck != attributeList.size()) {
-                LogError("ARFF/Integrity", "Value size mismatch: "
+                debug::Log(kError, "Value size mismatch: "
                         + std::to_string(successCheck) + " out of " 
                         + std::to_string(attributeList.size()));
                 exit(1);
             }
         }
-        LogInfo("ARFF/Integrity", "All values exist, continuing...");
+        debug::Log(kLog, "All values exist, continuing...");
+    }
+
+    // Perform OneR on data that was previously read in
+    AttributeEvaluation Arff::_OneR(void) {
+        AttributeEvaluation bestEvaluation;
+        bestEvaluation.totalErrorRate = 1.0f;
+        // -1 used for ignoring test rule (eg, play=yes/no)
+        for (int i = 0; i < attributeList.size() - 1; ++i) {
+            AttributeEvaluation evaluation = EvaluateAttribute(&attributeList[i], i);
+            if (evaluation.totalErrorRate < bestEvaluation.totalErrorRate) {
+                bestEvaluation = evaluation;
+                bestEvaluation.currentAttribute = evaluation.currentAttribute;
+            }
+            debug::Log(kLog, "Evaluation on " + evaluation.currentAttribute->attribute + " completed");
+        }
+        return bestEvaluation;
+    }
+
+    // Determine error rate and best option for each value of an attribute
+    // Originally set up to use OneR
+    AttributeEvaluation Arff::EvaluateAttribute(AttributeType *attribute, const int attributePos) {
+        AttributeEvaluation evaluation(attribute);
+        std::map<std::string, int> results;
+        for (std::string value : attributeList.end()->values) {
+            results.emplace(value, 0);
+        }
+        for (int i = 0; i < attribute->values.size(); ++i) {
+            int resultsTotal = 0;
+            if (attribute->values[i] == "?") { continue; }
+            for (auto instance = database.begin(); instance != database.end(); ++instance) {
+                if (instance->values[attributePos] != attribute->values[i]) { continue; }
+                ++results[instance->values.back()];
+                resultsTotal++;
+            }
+            debug::Log(kTrace, "Results:");
+            for (auto it : results) { debug::Log(kTrace, "\t" + it.first + ": " + std::to_string(it.second)); }
+            float lowest;
+            float lowestRate = 1.0f;
+            std::string bestResult = results.begin()->first;
+            for (auto it = results.begin(); it != results.end(); ++it) {
+                if (((resultsTotal - it->second) / float(resultsTotal)) < lowestRate) {
+                    lowestRate = ((resultsTotal - it->second) / float(resultsTotal));
+                    lowest = resultsTotal - it->second;
+                    bestResult = it->first;
+                } 
+            }
+            evaluation.rules.emplace(attribute->values[i], bestResult);
+            evaluation.totalError += lowest;
+            debug::Log(kLog, "Added rule " + attribute->values[i] + "->" + bestResult);
+            // Reset
+            for (auto it = results.begin(); it != results.end(); ++it) {
+                it->second = 0;
+            }
+        }
+        evaluation.totalErrorRate = evaluation.totalError / float(database.size());
+        return evaluation;
     }
 }

src/arff/arff.hpp

@@ -3,12 +3,11 @@
 
 #include <string>
 #include <vector>
+#include <map>
 
 namespace ARFF {
     void ParseArguments(int argc, char* argv[]);
     std::string GetDataFilename(void);
-    void LogInfo(const std::string location, const std::string message);
-    void LogError(const std::string location, const std::string message);
 
     struct AttributeType {
     public:
@@ -25,11 +24,22 @@ namespace ARFF {
         std::vector<std::string> values;
     };
 
+    struct AttributeEvaluation {
+        AttributeEvaluation() = default;
+        AttributeEvaluation(AttributeType *attribute);
+        AttributeType *currentAttribute;
+        std::map<std::string, std::string> rules;
+        float totalErrorRate = 0.0f;
+        int totalError = 0;
+    };
+
     class Arff {
     public:
         Arff() = default;
         void Read(std::string filename);
-        void Print(void);
+        void PrintOverview(void);
+        void PrintData(void);
+        void OneR(void);
     private:
         std::string relation;
         std::vector<AttributeType> attributeList;
@@ -37,6 +47,8 @@ namespace ARFF {
         void AddAttribute(std::string line);
         void AddData(std::string line);
         void TestIntegrity(void);
+        AttributeEvaluation _OneR(void);
+        AttributeEvaluation EvaluateAttribute(AttributeType *attribute, const int attributePos);
     };
 
 }

src/arff/log.cpp

@@ -0,0 +1,19 @@
+#include "log.hpp"
+#include <iostream>
+
+namespace debug {
+    bool verbose = false;
+
+    void Log(LogLevel level, std::string message) {
+        std::string logMessage = "";
+        if (!verbose && level > kNone) { return; }
+        switch (level) {
+            case kLog: logMessage += "[LOG] "; break;
+            case kWarn: logMessage += "[WARN] "; break;
+            case kError: logMessage += "[ERROR] "; break;
+            case kTrace: logMessage += "[TRACE] "; break;
+        }
+        logMessage += message;
+        std::cout << logMessage << std::endl;
+    }
+}

src/arff/log.hpp

@@ -0,0 +1,19 @@
+#ifndef LOG_HPP
+#define LOG_HPP
+
+#include <string>
+
+enum LogLevel {
+    kNone = -1,
+    kLog = 0,
+    kWarn,
+    kError,
+    kTrace
+};
+
+namespace debug {
+    extern bool verbose;
+    void Log(LogLevel level, std::string message);
+}
+
+#endif

src/arff/main.cpp

@@ -9,5 +9,6 @@ int main(int argc, char* argv[]) {
     ARFF::ParseArguments(argc, argv);
     ARFF::Arff data;
     data.Read(ARFF::GetDataFilename());
-    data.Print();
+    data.PrintOverview();
+    data.PrintData();
 }

src/onerule/CMakeLists.txt

@@ -0,0 +1,7 @@
+include_directories(../arff)
+add_executable(onerule
+    main.cpp
+    ../arff/arff.cpp
+    ../arff/log.cpp
+)
+

src/onerule/main.cpp

@@ -0,0 +1,14 @@
+/*
+ * Author: Gregory Crawford
+ * Date: 2024-03-18
+ * Description: Read data from ARFF file and generate OneR output 
+ */
+#include "arff.hpp"
+
+int main(int argc, char* argv[]) {
+    ARFF::ParseArguments(argc, argv);
+    ARFF::Arff data;
+    data.Read(ARFF::GetDataFilename());
+    data.PrintOverview();
+    data.OneR();
+}