**基于JStarCraft实现的搜索引擎** ## 1.项目介绍 **JStarCraft RNS是一个面向信息检索领域的轻量级引擎.遵循Apache 2.0协议.** 专注于解决信息检索领域的基本问题:推荐与搜索. 提供满足工业级别场景要求的推荐引擎设计与实现. 提供满足工业级别场景要求的搜索引擎设计与实现. **** ## 2.特性 * 1.跨平台 * [2.串行与并行计算](https://github.com/HongZhaoHua/jstarcraft-ai) * [3.CPU与GPU硬件加速](https://github.com/HongZhaoHua/jstarcraft-ai) * [4.模型保存与装载](https://github.com/HongZhaoHua/jstarcraft-ai) * 5.丰富的推荐与搜索算法 * 6.丰富的脚本支持 * Groovy * JS * Lua * MVEL * Python * Ruby * [7.丰富的评估指标](#评估指标) * [排序指标](#排序指标) * [评分指标](#评分指标) **** ## 3.安装 JStarCraft RNS要求使用者具备以下环境: * JDK 8或者以上 * Maven 3 #### 3.1安装JStarCraft-Core框架 ```shell git clone https://github.com/HongZhaoHua/jstarcraft-core.git mvn install -Dmaven.test.skip=true ``` #### 3.2安装JStarCraft-AI框架 ```shell git clone https://github.com/HongZhaoHua/jstarcraft-ai.git mvn install -Dmaven.test.skip=true ``` #### 3.3安装JStarCraft-RNS引擎 ```shell git clone https://github.com/HongZhaoHua/jstarcraft-rns.git mvn install -Dmaven.test.skip=true ``` **** ## 4.使用 #### 4.1设置依赖 * 设置Maven依赖 ```xml com.jstarcraft rns 1.0 ``` * 设置Gradle依赖 ```gradle compile group: 'com.jstarcraft', name: 'rns', version: '1.0' ``` #### 4.2构建配置器 ```java Properties keyValues = new Properties(); keyValues.load(this.getClass().getResourceAsStream("/data.properties")); keyValues.load(this.getClass().getResourceAsStream("/recommend/benchmark/randomguess-test.properties")); Configurator configurator = new Configurator(keyValues); ``` #### 4.3训练与评估模型 * 构建排序任务 ```java RankingTask task = new RankingTask(RandomGuessModel.class, configurator); // 训练与评估排序模型 task.execute(); ``` * 构建评分任务 ```java RatingTask task = new RatingTask(RandomGuessModel.class, configurator); // 训练与评估评分模型 task.execute(); ``` #### 4.4获取模型 ```java // 获取模型 Model model = task.getModel(); ``` **** ## 5.概念 #### 5.1为什么需要信息检索 ``` 随着信息技术和互联网的发展,人们逐渐从信息匮乏(Information Underload)的时代走入了信息过载(Information Overload)的时代. 无论是信息消费者还是信息生产者都遇到了挑战: * 对于信息消费者,从海量信息中寻找信息,是一件非常困难的事情; * 对于信息生产者,从海量信息中暴露信息,也是一件非常困难的事情; 信息检索的任务就是联系用户和信息,一方面帮助用户寻找对自己有价值的信息,另一方面帮助信息暴露给对它感兴趣的用户,从而实现信息消费者和信息生产者的双赢. ``` #### 5.2搜索与推荐的异同 ``` 从信息检索的角度: * 搜索和推荐是获取信息的两种主要手段; * 搜索和推荐是获取信息的两种不同方式; * 搜索(Search)是主动明确的; * 推荐(Recommend)是被动模糊的; 搜索和推荐是两个互补的工具. ``` #### 5.3JStarCraft-RNS引擎解决什么问题 ``` JStarCraft-RNS引擎旨在解决推荐与搜索领域的两个核心任务:排序预测(Ranking)和评分预测(Rating). ``` #### 5.4Ranking任务与Rating任务之间的区别 ``` 根据解决基本问题的不同,将算法与评估指标划分为排序(Ranking)与评分(Rating). 两者之间的根本区别在于目标函数的不同. 通俗点的解释: Ranking算法基于隐式反馈数据,趋向于拟合用户的排序.(关注度) Rating算法基于显示反馈数据,趋向于拟合用户的评分.(满意度) ``` #### 5.5Rating算法能不能用于Ranking问题 ``` 关键在于具体场景中,关注度与满意度是否保持一致. 通俗点的解释: 人们关注的东西,并不一定是满意的东西.(例如:个人所得税) ``` **** ## 6.示例 #### 6.1JStarCraft-RNS引擎与BeanShell脚本交互 * [完整示例](https://github.com/HongZhaoHua/jstarcraft-rns/tree/master/src/test/java/com/jstarcraft/rns/script) * 编写BeanShell脚本训练与评估模型并保存到Model.bsh文件 ```java // 构建配置 keyValues = new Properties(); keyValues.load(loader.getResourceAsStream("data.properties")); keyValues.load(loader.getResourceAsStream("model/benchmark/randomguess-test.properties")); configurator = new Configurator(keyValues); // 此对象会返回给Java程序 _data = new HashMap(); // 构建排序任务 task = new RankingTask(RandomGuessModel.class, configurator); // 训练与评估模型并获取排序指标 measures = task.execute(); _data.put("precision", measures.get(PrecisionEvaluator.class)); _data.put("recall", measures.get(RecallEvaluator.class)); // 构建评分任务 task = new RatingTask(RandomGuessModel.class, configurator); // 训练与评估模型并获取评分指标 measures = task.execute(); _data.put("mae", measures.get(MAEEvaluator.class)); _data.put("mse", measures.get(MSEEvaluator.class)); _data; ``` * 使用JStarCraft框架从Model.bsh文件加载并执行BeanShell脚本 ```java // 获取BeanShell脚本 File file = new File(ScriptTestCase.class.getResource("Model.bsh").toURI()); String script = FileUtils.readFileToString(file, StringUtility.CHARSET); // 设置BeanShell脚本使用到的Java类 ScriptContext context = new ScriptContext(); context.useClasses(Properties.class, Assert.class); context.useClass("Configurator", MapConfigurator.class); context.useClasses("com.jstarcraft.ai.evaluate"); context.useClasses("com.jstarcraft.rns.task"); context.useClasses("com.jstarcraft.rns.model.benchmark"); // 设置BeanShell脚本使用到的Java变量 ScriptScope scope = new ScriptScope(); scope.createAttribute("loader", loader); // 执行BeanShell脚本 ScriptExpression expression = new GroovyExpression(context, scope, script); Map data = expression.doWith(Map.class); Assert.assertEquals(0.005825241F, data.get("precision"), 0F); Assert.assertEquals(0.011579763F, data.get("recall"), 0F); Assert.assertEquals(1.2708743F, data.get("mae"), 0F); Assert.assertEquals(2.425075F, data.get("mse"), 0F); ``` #### 6.2JStarCraft-RNS引擎与Groovy脚本交互 * [完整示例](https://github.com/HongZhaoHua/jstarcraft-rns/tree/master/src/test/java/com/jstarcraft/rns/script) * 编写Groovy脚本训练与评估模型并保存到Model.groovy文件 ```groovy // 构建配置 def keyValues = new Properties(); keyValues.load(loader.getResourceAsStream("data.properties")); keyValues.load(loader.getResourceAsStream("recommend/benchmark/randomguess-test.properties")); def configurator = new Configurator(keyValues); // 此对象会返回给Java程序 def _data = [:]; // 构建排序任务 task = new RankingTask(RandomGuessModel.class, configurator); // 训练与评估模型并获取排序指标 measures = task.execute(); _data.precision = measures.get(PrecisionEvaluator.class); _data.recall = measures.get(RecallEvaluator.class); // 构建评分任务 task = new RatingTask(RandomGuessModel.class, configurator); // 训练与评估模型并获取评分指标 measures = task.execute(); _data.mae = measures.get(MAEEvaluator.class); _data.mse = measures.get(MSEEvaluator.class); _data; ``` * 使用JStarCraft框架从Model.groovy文件加载并执行Groovy脚本 ```java // 获取Groovy脚本 File file = new File(ScriptTestCase.class.getResource("Model.groovy").toURI()); String script = FileUtils.readFileToString(file, StringUtility.CHARSET); // 设置Groovy脚本使用到的Java类 ScriptContext context = new ScriptContext(); context.useClasses(Properties.class, Assert.class); context.useClass("Configurator", MapConfigurator.class); context.useClasses("com.jstarcraft.ai.evaluate"); context.useClasses("com.jstarcraft.rns.task"); context.useClasses("com.jstarcraft.rns.model.benchmark"); // 设置Groovy脚本使用到的Java变量 ScriptScope scope = new ScriptScope(); scope.createAttribute("loader", loader); // 执行Groovy脚本 ScriptExpression expression = new GroovyExpression(context, scope, script); Map data = expression.doWith(Map.class); ``` #### 6.3JStarCraft-RNS引擎与JS脚本交互 * [完整示例](https://github.com/HongZhaoHua/jstarcraft-rns/tree/master/src/test/java/com/jstarcraft/rns/script) * 编写JS脚本训练与评估模型并保存到Model.js文件 ```js // 构建配置 var keyValues = new Properties(); keyValues.load(loader.getResourceAsStream("data.properties")); keyValues.load(loader.getResourceAsStream("recommend/benchmark/randomguess-test.properties")); var configurator = new Configurator([keyValues]); // 此对象会返回给Java程序 var _data = {}; // 构建排序任务 task = new RankingTask(RandomGuessModel.class, configurator); // 训练与评估模型并获取排序指标 measures = task.execute(); _data['precision'] = measures.get(PrecisionEvaluator.class); _data['recall'] = measures.get(RecallEvaluator.class); // 构建评分任务 task = new RatingTask(RandomGuessModel.class, configurator); // 训练与评估模型并获取评分指标 measures = task.execute(); _data['mae'] = measures.get(MAEEvaluator.class); _data['mse'] = measures.get(MSEEvaluator.class); _data; ``` * 使用JStarCraft框架从Model.js文件加载并执行JS脚本 ```java // 获取JS脚本 File file = new File(ScriptTestCase.class.getResource("Model.js").toURI()); String script = FileUtils.readFileToString(file, StringUtility.CHARSET); // 设置JS脚本使用到的Java类 ScriptContext context = new ScriptContext(); context.useClasses(Properties.class, Assert.class); context.useClass("Configurator", MapConfigurator.class); context.useClasses("com.jstarcraft.ai.evaluate"); context.useClasses("com.jstarcraft.rns.task"); context.useClasses("com.jstarcraft.rns.model.benchmark"); // 设置JS脚本使用到的Java变量 ScriptScope scope = new ScriptScope(); scope.createAttribute("loader", loader); // 执行JS脚本 ScriptExpression expression = new JsExpression(context, scope, script); Map data = expression.doWith(Map.class); ``` #### 6.4JStarCraft-RNS引擎与Kotlin脚本交互 * [完整示例](https://github.com/HongZhaoHua/jstarcraft-rns/tree/master/src/test/java/com/jstarcraft/rns/script) * 编写Kotlin脚本训练与评估模型并保存到Model.kt文件 ```js // 构建配置 var keyValues = Properties(); var loader = bindings["loader"] as ClassLoader; keyValues.load(loader.getResourceAsStream("data.properties")); keyValues.load(loader.getResourceAsStream("model/benchmark/randomguess-test.properties")); var option = Option(keyValues); // 此对象会返回给Java程序 var _data = mutableMapOf(); // 构建排序任务 var rankingTask = RankingTask(RandomGuessModel::class.java, option); // 训练与评估模型并获取排序指标 val rankingMeasures = rankingTask.execute(); _data["precision"] = rankingMeasures.getFloat(PrecisionEvaluator::class.java); _data["recall"] = rankingMeasures.getFloat(RecallEvaluator::class.java); // 构建评分任务 var ratingTask = RatingTask(RandomGuessModel::class.java, option); // 训练与评估模型并获取评分指标 var ratingMeasures = ratingTask.execute(); _data["mae"] = ratingMeasures.getFloat(MAEEvaluator::class.java); _data["mse"] = ratingMeasures.getFloat(MSEEvaluator::class.java); _data; ``` * 使用JStarCraft框架从Model.kt文件加载并执行Kotlin脚本 ```java // 获取Kotlin脚本 File file = new File(ScriptTestCase.class.getResource("Model.kt").toURI()); String script = FileUtils.readFileToString(file, StringUtility.CHARSET); // 设置Kotlin脚本使用到的Java类 ScriptContext context = new ScriptContext(); context.useClasses(Properties.class, Assert.class); context.useClass("Option", MapOption.class); context.useClasses("com.jstarcraft.ai.evaluate"); context.useClasses("com.jstarcraft.rns.task"); context.useClasses("com.jstarcraft.rns.model.benchmark"); // 设置Kotlin脚本使用到的Java变量 ScriptScope scope = new ScriptScope(); scope.createAttribute("loader", loader); // 执行Kotlin脚本 ScriptExpression expression = new KotlinExpression(context, scope, script); Map data = expression.doWith(Map.class); ``` #### 6.5JStarCraft-RNS引擎与Lua脚本交互 * [完整示例](https://github.com/HongZhaoHua/jstarcraft-rns/tree/master/src/test/java/com/jstarcraft/rns/script) * 编写Lua脚本训练与评估模型并保存到Model.lua文件 ```lua -- 构建配置 local keyValues = Properties.new(); keyValues:load(loader:getResourceAsStream("data.properties")); keyValues:load(loader:getResourceAsStream("recommend/benchmark/randomguess-test.properties")); local configurator = Configurator.new({ keyValues }); -- 此对象会返回给Java程序 local _data = {}; -- 构建排序任务 task = RankingTask.new(RandomGuessModel, configurator); -- 训练与评估模型并获取排序指标 measures = task:execute(); _data["precision"] = measures:get(PrecisionEvaluator); _data["recall"] = measures:get(RecallEvaluator); -- 构建评分任务 task = RatingTask.new(RandomGuessModel, configurator); -- 训练与评估模型并获取评分指标 measures = task:execute(); _data["mae"] = measures:get(MAEEvaluator); _data["mse"] = measures:get(MSEEvaluator); return _data; ``` * 使用JStarCraft框架从Model.lua文件加载并执行Lua脚本 ```java // 获取Lua脚本 File file = new File(ScriptTestCase.class.getResource("Model.lua").toURI()); String script = FileUtils.readFileToString(file, StringUtility.CHARSET); // 设置Lua脚本使用到的Java类 ScriptContext context = new ScriptContext(); context.useClasses(Properties.class, Assert.class); context.useClass("Configurator", MapConfigurator.class); context.useClasses("com.jstarcraft.ai.evaluate"); context.useClasses("com.jstarcraft.rns.task"); context.useClasses("com.jstarcraft.rns.model.benchmark"); // 设置Lua脚本使用到的Java变量 ScriptScope scope = new ScriptScope(); scope.createAttribute("loader", loader); // 执行Lua脚本 ScriptExpression expression = new LuaExpression(context, scope, script); LuaTable data = expression.doWith(LuaTable.class); ``` #### 6.6JStarCraft-RNS引擎与Python脚本交互 * [完整示例](https://github.com/HongZhaoHua/jstarcraft-rns/tree/master/src/test/java/com/jstarcraft/rns/script) * 编写Python脚本训练与评估模型并保存到Model.py文件 ```python # 构建配置 keyValues = Properties() keyValues.load(loader.getResourceAsStream("data.properties")) keyValues.load(loader.getResourceAsStream("recommend/benchmark/randomguess-test.properties")) configurator = Configurator([keyValues]) # 此对象会返回给Java程序 _data = {} # 构建排序任务 task = RankingTask(RandomGuessModel, configurator) # 训练与评估模型并获取排序指标 measures = task.execute() _data['precision'] = measures.get(PrecisionEvaluator) _data['recall'] = measures.get(RecallEvaluator) # 构建评分任务 task = RatingTask(RandomGuessModel, configurator) # 训练与评估模型并获取评分指标 measures = task.execute() _data['mae'] = measures.get(MAEEvaluator) _data['mse'] = measures.get(MSEEvaluator) ``` * 使用JStarCraft框架从Model.py文件加载并执行Python脚本 ```java // 设置Python环境变量 System.setProperty("python.console.encoding", StringUtility.CHARSET.name()); // 获取Python脚本 File file = new File(PythonTestCase.class.getResource("Model.py").toURI()); String script = FileUtils.readFileToString(file, StringUtility.CHARSET); // 设置Python脚本使用到的Java类 ScriptContext context = new ScriptContext(); context.useClasses(Properties.class, Assert.class); context.useClass("Configurator", MapConfigurator.class); context.useClasses("com.jstarcraft.ai.evaluate"); context.useClasses("com.jstarcraft.rns.task"); context.useClasses("com.jstarcraft.rns.model.benchmark"); // 设置Python脚本使用到的Java变量 ScriptScope scope = new ScriptScope(); scope.createAttribute("loader", loader); // 执行Python脚本 ScriptExpression expression = new PythonExpression(context, scope, script); Map data = expression.doWith(Map.class); ``` #### 6.7JStarCraft-Ruby * [完整示例](https://github.com/HongZhaoHua/jstarcraft-rns/tree/master/src/test/java/com/jstarcraft/rns/script) * 编写Ruby脚本训练与评估模型并保存到Model.rb文件 ```ruby # 构建配置 keyValues = Properties.new() keyValues.load($loader.getResourceAsStream("data.properties")) keyValues.load($loader.getResourceAsStream("model/benchmark/randomguess-test.properties")) configurator = Configurator.new(keyValues) # 此对象会返回给Java程序 _data = Hash.new() # 构建排序任务 task = RankingTask.new(RandomGuessModel.java_class, configurator) # 训练与评估模型并获取排序指标 measures = task.execute() _data['precision'] = measures.get(PrecisionEvaluator.java_class) _data['recall'] = measures.get(RecallEvaluator.java_class) # 构建评分任务 task = RatingTask.new(RandomGuessModel.java_class, configurator) # 训练与评估模型并获取评分指标 measures = task.execute() _data['mae'] = measures.get(MAEEvaluator.java_class) _data['mse'] = measures.get(MSEEvaluator.java_class) _data; ``` * 使用JStarCraft框架从Model.rb文件加载并执行Ruby脚本 ```java // 获取Ruby脚本 File file = new File(ScriptTestCase.class.getResource("Model.rb").toURI()); String script = FileUtils.readFileToString(file, StringUtility.CHARSET); // 设置Ruby脚本使用到的Java类 ScriptContext context = new ScriptContext(); context.useClasses(Properties.class, Assert.class); context.useClass("Configurator", MapConfigurator.class); context.useClasses("com.jstarcraft.ai.evaluate"); context.useClasses("com.jstarcraft.rns.task"); context.useClasses("com.jstarcraft.rns.model.benchmark"); // 设置Ruby脚本使用到的Java变量 ScriptScope scope = new ScriptScope(); scope.createAttribute("loader", loader); // 执行Ruby脚本 ScriptExpression expression = new RubyExpression(context, scope, script); Map data = expression.doWith(Map.class); Assert.assertEquals(0.005825241096317768D, data.get("precision"), 0D); Assert.assertEquals(0.011579763144254684D, data.get("recall"), 0D); Assert.assertEquals(1.270874261856079D, data.get("mae"), 0D); Assert.assertEquals(2.425075054168701D, data.get("mse"), 0D); ``` **** ## 7.对比 #### 7.1排序模型对比 * 基准模型 | 名称 | 数据集 | 训练 (毫秒) | 预测 (毫秒) | AUC | MAP | MRR | NDCG | Novelty | Precision | Recall | | :----: | :----: | :----: | :----: | :----: | :----: | :----: | :----: | :----: | :----: | :----: | | MostPopular | filmtrust | 43 | 273 | 0.92080 | 0.41246 | 0.57196 | 0.51583 | 11.79295 | 0.33230 | 0.62385 | | RandomGuess | filmtrust | 38 | 391 | 0.51922 | 0.00627 | 0.02170 | 0.01121 | 91.94900 | 0.00550 | 0.01262 | * 协同模型 | 名称 | 数据集 | 训练 (毫秒) | 预测 (毫秒) | AUC | MAP | MRR | NDCG | Novelty | Precision | Recall | | :----: | :----: | :----: | :----: | :----: | :----: | :----: | :----: | :----: | :----: | :----: | | AoBPR | filmtrust | 12448 | 253 | 0.89324 | 0.38967 | 0.53990 | 0.48338 | 21.13004 | 0.32295 | 0.56864 | | AspectRanking | filmtrust | 177 | 58 | 0.85130 | 0.15498 | 0.42480 | 0.26012 | 37.36273 | 0.13302 | 0.31292 | | BHFreeRanking | filmtrust | 5720 | 4257 | 0.92080 | 0.41316 | 0.57231 | 0.51662 | 11.79567 | 0.33276 | 0.62500 | | BPR | filmtrust | 4228 | 137 | 0.89390 | 0.39886 | 0.54790 | 0.49180 | 21.46738 | 0.32268 | 0.57623 | | BUCMRanking | filmtrust | 2111 | 1343 | 0.90782 | 0.39794 | 0.55776 | 0.49651 | 13.08073 | 0.32407 | 0.59141 | | CDAE | filmtrust | 89280 | 376 | 0.91880 | 0.40759 | 0.56855 | 0.51089 | 11.82466 | 0.33051 | 0.61967 | | CLiMF | filmtrust | 48429 | 140 | 0.88293 | 0.37395 | 0.52407 | 0.46572 | 19.38964 | 0.32049 | 0.54605 | | DeepFM | filmtrust | 69264 | 99 | 0.91679 | 0.40580 | 0.56995 | 0.50985 | 11.90242 | 0.32719 | 0.61426 | | EALS | filmtrust | 850 | 185 | 0.86132 | 0.31263 | 0.45680 | 0.39475 | 20.08964 | 0.27381 | 0.46271 | | FISMAUC | filmtrust | 2338 | 663 | 0.91216 | 0.40032 | 0.55730 | 0.50114 | 12.07469 | 0.32845 | 0.60294 | | FISMRMSE | filmtrust | 4030 | 729 | 0.91482 | 0.40795 | 0.56470 | 0.50920 | 11.91234 | 0.33044 | 0.61107 | | GBPR | filmtrust | 14827 | 150 | 0.92113 | 0.41003 | 0.57144 | 0.51464 | 11.87609 | 0.33090 | 0.62512 | | HMM | game | 38697 | 11223 | 0.80559 | 0.18156 | 0.37516 | 0.25803 | 16.01041 | 0.14572 | 0.22810 | | ItemBigram | filmtrust | 12492 | 61 | 0.88807 | 0.33520 | 0.46870 | 0.42854 | 17.11172 | 0.29191 | 0.53308 | | ItemKNNRanking | filmtrust | 2683 | 250 | 0.87438 | 0.33375 | 0.46951 | 0.41767 | 20.23449 | 0.28581 | 0.49248 | | LDA | filmtrust | 696 | 161 | 0.91980 | 0.41758 | 0.58130 | 0.52003 | 12.31348 | 0.33336 | 0.62274 | | LambdaFMStatic | game | 25052 | 27078 | 0.87064 | 0.27294 | 0.43640 | 0.34794 | 16.47330 | 0.13941 | 0.35696 | | LambdaFMWeight | game | 25232 | 28156 | 0.87339 | 0.27333 | 0.43720 | 0.34728 | 14.71413 | 0.13742 | 0.35252 | | LambdaFMDynamic | game | 74218 | 27921 | 0.87380 | 0.27288 | 0.43648 | 0.34706 | 13.50578 | 0.13822 | 0.35132 | | ListwiseMF | filmtrust | 714 | 161 | 0.90820 | 0.40511 | 0.56619 | 0.50521 | 15.53665 | 0.32944 | 0.60092 | | PLSA | filmtrust | 1027 | 116 | 0.89950 | 0.41217 | 0.57187 | 0.50597 | 16.01080 | 0.32401 | 0.58557 | | RankALS | filmtrust | 3285 | 182 | 0.85901 | 0.29255 | 0.51014 | 0.38871 | 25.27197 | 0.22931 | 0.42509 | | RankCD | product | 1442 | 8905 | 0.56271 | 0.01253 | 0.04618 | 0.02682 | 55.42019 | 0.01548 | 0.03520 | | RankSGD | filmtrust | 309 | 113 | 0.80388 | 0.23587 | 0.42290 | 0.32081 | 42.83305 | 0.19363 | 0.35374 | | RankVFCD | product | 54273 | 6524 | 0.58022 | 0.01784 | 0.06181 | 0.03664 | 62.95810 | 0.01980 | 0.04852 | | SLIM | filmtrust | 62434 | 91 | 0.91849 | 0.44851 | 0.61083 | 0.54557 | 16.67990 | 0.34019 | 0.63021 | | UserKNNRanking | filmtrust | 1154 | 229 | 0.90752 | 0.41616 | 0.57525 | 0.51393 | 12.90921 | 0.32891 | 0.60152 | | VBPR | product | 184473 | 15304 | 0.54336 | 0.00920 | 0.03522 | 0.01883 | 45.05101 | 0.01037 | 0.02266 | | WBPR | filmtrust | 20705 | 183 | 0.78072 | 0.24647 | 0.33373 | 0.30442 | 17.18609 | 0.25000 | 0.35516 | | WRMF | filmtrust | 482 | 158 | 0.90616 | 0.43278 | 0.58284 | 0.52480 | 15.17956 | 0.32918 | 0.60780 | | RankGeoFM | FourSquare | 368436 | 1093 | 0.72708 | 0.05485 | 0.24012 | 0.11057 | 37.50040 | 0.07866 | 0.08640 | | SBPR | filmtrust | 41481 | 247 | 0.91010 | 0.41189 | 0.56480 | 0.50726 | 15.67905 | 0.32440 | 0.59699 | * 内容模型 | 名称 | 数据集 | 训练 (毫秒) | 预测 (毫秒) | AUC | MAP | MRR | NDCG | Novelty | Precision | Recall | | :----: | :----: | :----: | :----: | :----: | :----: | :----: | :----: | :----: | :----: | :----: | | EFMRanking | dc_dense | 2066 | 2276 | 0.61271 | 0.01611 | 0.04631 | 0.04045 | 53.26140 | 0.02387 | 0.07357 | | TFIDF | musical_instruments | 942 | 1085 | 0.52756 | 0.01067 | 0.01917 | 0.01773 | 72.71228 | 0.00588 | 0.03103 | #### 7.2评分模型对比 * 基准模型 | 名称 | 数据集 | 训练 (毫秒) | 预测 (毫秒) | MAE | MPE | MSE | | :----: | :----: | :----: | :----: | :----: | :----: | :----: | | ConstantGuess | filmtrust | 137 | 45 | 1.05608 | 1.00000 | 1.42309 | | GlobalAverage | filmtrust | 60 | 13 | 0.71977 | 0.77908 | 0.85199 | | ItemAverage | filmtrust | 59 | 12 | 0.72968 | 0.97242 | 0.86413 | | ItemCluster | filmtrust | 471 | 41 | 0.71976 | 0.77908 | 0.85198 | | RandomGuess | filmtrust | 38 | 8 | 1.28622 | 0.99597 | 2.47927 | | UserAverage | filmtrust | 35 | 9 | 0.64618 | 0.97242 | 0.70172 | | UserCluster | filmtrust | 326 | 45 | 0.71977 | 0.77908 | 0.85199 | * 协同模型 | 名称 | 数据集 | 训练 (毫秒) | 预测 (毫秒) | MAE | MPE | MSE | | :----: | :----: | :----: | :----: | :----: | :----: | :----: | | AspectRating | filmtrust | 220 | 5 | 0.65754 | 0.97918 | 0.71809 | | ASVDPlusPlus | filmtrust | 5631 | 8 | 0.71975 | 0.77921 | 0.85196 | | BiasedMF | filmtrust | 92 | 6 | 0.63157 | 0.98387 | 0.66220 | | BHFreeRating | filmtrust | 6667 | 76 | 0.71974 | 0.77908 | 0.85198 | | BPMF | filmtrust | 25942 | 52 | 0.66504 | 0.98465 | 0.70210 | | BUCMRating | filmtrust | 1843 | 30 | 0.64834 | 0.99102 | 0.67992 | | CCD | product | 15715 | 9 | 0.96670 | 0.93947 | 1.62145 | | FFM | filmtrust | 5422 | 6 | 0.63446 | 0.98413 | 0.66682 | | FMALS | filmtrust | 1854 | 5 | 0.64788 | 0.96032 | 0.73636 | | FMSGD | filmtrust | 3496 | 10 | 0.63452 | 0.98426 | 0.66710 | | GPLSA | filmtrust | 2567 | 7 | 0.67311 | 0.98972 | 0.79883 | | IRRG | filmtrust | 40284 | 6 | 0.64766 | 0.98777 | 0.73700 | | ItemKNNRating | filmtrust | 2052 | 27 | 0.62341 | 0.95394 | 0.67312 | | LDCC | filmtrust | 8650 | 84 | 0.66383 | 0.99284 | 0.70666 | | LLORMA | filmtrust | 16618 | 82 | 0.64930 | 0.96591 | 0.76067 | | MFALS | filmtrust | 2944 | 5 | 0.82939 | 0.94549 | 1.30547 | | NMF | filmtrust | 1198 | 8 | 0.67661 | 0.96604 | 0.83493 | | PMF | filmtrust | 215 | 7 | 0.72959 | 0.98165 | 0.99948 | | RBM | filmtrust | 19551 | 270 | 0.74484 | 0.98504 | 0.88968 | | RFRec | filmtrust | 16330 | 54 | 0.64008 | 0.97112 | 0.69390 | | SVDPlusPlus | filmtrust | 452 | 26 | 0.65248 | 0.99141 | 0.68289 | | URP | filmtrust | 1514 | 25 | 0.64207 | 0.99128 | 0.67122 | | UserKNNRating | filmtrust | 1121 | 135 | 0.63933 | 0.94640 | 0.69280 | | RSTE | filmtrust | 4052 | 10 | 0.64303 | 0.99206 | 0.67777 | | SocialMF | filmtrust | 918 | 13 | 0.64668 | 0.98881 | 0.68228 | | SoRec | filmtrust | 1048 | 10 | 0.64305 | 0.99232 | 0.67776 | | SoReg | filmtrust | 635 | 8 | 0.65943 | 0.96734 | 0.72760 | | TimeSVD | filmtrust | 11545 | 36 | 0.68954 | 0.93326 | 0.87783 | | TrustMF | filmtrust | 2038 | 7 | 0.63787 | 0.98985 | 0.69017 | | TrustSVD | filmtrust | 12465 | 22 | 0.61984 | 0.98933 | 0.63875 | | AssociationRule | filmtrust | 2628 | 195 | 0.90853 | 0.41801 | 0.57777 | 0.51621 | 12.65794 | 0.33263 | 0.60700 | | PersonalityDiagnosis | filmtrust | 45 | 642 | 0.72964 | 0.76620 | 1.03071 | | PRankD | filmtrust | 3321 | 170 | 0.74472 | 0.22894 | 0.32406 | 0.28390 | 45.81069 | 0.19436 | 0.32904 | | SlopeOne | filmtrust | 135 | 28 | 0.63788 | 0.96175 | 0.71057 | * 内容模型 | 名称 | 数据集 | 训练 (毫秒) | 预测 (毫秒) | MAE | MPE | MSE | | :----: | :----: | :----: | :----: | :----: | :----: | :----: | | EFMRating | dc_dense | 659 | 8 | 0.61546 | 0.85364 | 0.78279 | | HFT | musical_instruments | 162753 | 13 | 0.64272 | 0.94886 | 0.81393 | | TopicMFAT | musical_instruments | 6907 | 7 | 0.61896 | 0.98734 | 0.72545 | | TopicMFMT | musical_instruments | 6323 | 7 | 0.61896 | 0.98734 | 0.72545 | ## 8.参考 #### 8.1个性化模型说明 * 基准模型 | 名称 | 问题 | 说明/论文 | | :----: | :----: | :----: | | RandomGuess | Ranking Rating | 随机猜测 | | MostPopular | Ranking| 最受欢迎 | | ConstantGuess | Rating | 常量猜测 | | GlobalAverage | Rating | 全局平均 | | ItemAverage | Rating | 物品平均 | | ItemCluster | Rating | 物品聚类 | | UserAverage | Rating | 用户平均 | | UserCluster | Rating | 用户聚类 | * 协同模型 | 名称 | 问题 | 说明/论文 | | :----: | :----: | :----: | | AspectModel | Ranking Rating | Latent class models for collaborative filtering | | BHFree | Ranking Rating | Balancing Prediction and Recommendation Accuracy: Hierarchical Latent Factors for Preference Data | | BUCM | Ranking Rating | Modeling Item Selection and Relevance for Accurate Recommendations | | ItemKNN | Ranking Rating | 基于物品的协同过滤 | | UserKNN | Ranking Rating | 基于用户的协同过滤 | | AoBPR | Ranking | Improving pairwise learning for item recommendation from implicit feedback | | BPR | Ranking | BPR: Bayesian Personalized Ranking from Implicit Feedback | | CLiMF | Ranking | CLiMF: learning to maximize reciprocal rank with collaborative less-is-more filtering | | EALS | Ranking | Collaborative filtering for implicit feedback dataset | | FISM | Ranking | FISM: Factored Item Similarity Models for Top-N Recommender Systems | | GBPR | Ranking | GBPR: Group Preference Based Bayesian Personalized Ranking for One-Class Collaborative Filtering | | HMMForCF | Ranking | A Hidden Markov Model Purpose: A class for the model, including parameters | | ItemBigram | Ranking | Topic Modeling: Beyond Bag-of-Words | | LambdaFM | Ranking | LambdaFM: Learning Optimal Ranking with Factorization Machines Using Lambda Surrogates | | LDA | Ranking | Latent Dirichlet Allocation for implicit feedback | | ListwiseMF | Ranking | List-wise learning to rank with matrix factorization for collaborative filtering | | PLSA | Ranking | Latent semantic models for collaborative filtering | | RankALS | Ranking | Alternating Least Squares for Personalized Ranking | | RankSGD | Ranking | Collaborative Filtering Ensemble for Ranking | | SLIM | Ranking | SLIM: Sparse Linear Methods for Top-N Recommender Systems | | WBPR | Ranking | Bayesian Personalized Ranking for Non-Uniformly Sampled Items | | WRMF | Ranking | Collaborative filtering for implicit feedback datasets | | Rank-GeoFM | Ranking | Rank-GeoFM: A ranking based geographical factorization method for point of interest recommendation | | SBPR | Ranking | Leveraging Social Connections to Improve Personalized Ranking for Collaborative Filtering | | AssociationRule | Ranking | A Recommendation Algorithm Using Multi-Level Association Rules | | PRankD | Ranking | Personalised ranking with diversity | | AsymmetricSVD++ | Rating | Factorization Meets the Neighborhood: a Multifaceted Collaborative Filtering Model | | AutoRec | Rating | AutoRec: Autoencoders Meet Collaborative Filtering | | BPMF | Rating | Bayesian Probabilistic Matrix Factorization using Markov Chain Monte Carlo | | CCD | Rating | Large-Scale Parallel Collaborative Filtering for the Netflix Prize | | FFM | Rating | Field Aware Factorization Machines for CTR Prediction | | GPLSA | Rating | Collaborative Filtering via Gaussian Probabilistic Latent Semantic Analysis | | IRRG | Rating | Exploiting Implicit Item Relationships for Recommender Systems | | MFALS | Rating | Large-Scale Parallel Collaborative Filtering for the Netflix Prize | | NMF | Rating | Algorithms for Non-negative Matrix Factorization | | PMF | Rating | PMF: Probabilistic Matrix Factorization | | RBM | Rating | Restricted Boltzman Machines for Collaborative Filtering | | RF-Rec | Rating | RF-Rec: Fast and Accurate Computation of Recommendations based on Rating Frequencies | | SVD++ | Rating | Factorization Meets the Neighborhood: a Multifaceted Collaborative Filtering Model | | URP | Rating | User Rating Profile: a LDA model for rating prediction | | RSTE | Rating | Learning to Recommend with Social Trust Ensemble | | SocialMF | Rating | A matrix factorization technique with trust propagation for recommendation in social networks | | SoRec | Rating | SoRec: Social recommendation using probabilistic matrix factorization | | SoReg | Rating | Recommender systems with social regularization | | TimeSVD++ | Rating | Collaborative Filtering with Temporal Dynamics | | TrustMF | Rating | Social Collaborative Filtering by Trust | | TrustSVD | Rating | TrustSVD: Collaborative Filtering with Both the Explicit and Implicit Influence of User Trust and of Item Ratings | | PersonalityDiagnosis | Rating | A brief introduction to Personality Diagnosis | | SlopeOne | Rating | Slope One Predictors for Online Rating-Based Collaborative Filtering | * 内容模型 | 名称 | 问题 | 说明/论文 | | :----: | :----: | :----: | | EFM | Ranking Rating | Explicit factor models for explainable recommendation based on phrase-level sentiment analysis | | TF-IDF | Ranking | 词频-逆文档频率 | | HFT | Rating | Hidden factors and hidden topics: understanding rating dimensions with review text | | TopicMF | Rating | TopicMF: Simultaneously Exploiting Ratings and Reviews for Recommendation | #### 8.2数据集说明 * [Amazon Dataset](http://jmcauley.ucsd.edu/data/amazon/) * [Bibsonomy Dataset](https://www.kde.cs.uni-kassel.de/wp-content/uploads/bibsonomy/) * [BookCrossing Dataset](https://grouplens.org/datasets/book-crossing/) * [Ciao Dataset](https://www.cse.msu.edu/~tangjili/datasetcode/truststudy.htm) * [Douban Dataset](http://smiles.xjtu.edu.cn/Download/Download_Douban.html) * [Eachmovie Dataset](https://grouplens.org/datasets/eachmovie/) * [Epinions Dataset](http://www.trustlet.org/epinions.html) * [Foursquare Dataset](https://sites.google.com/site/yangdingqi/home/foursquare-dataset) * [Goodbooks Dataset](http://fastml.com/goodbooks-10k-a-new-dataset-for-book-recommendations/) * [Gowalla Dataset](http://snap.stanford.edu/data/loc-gowalla.html) * [HetRec2011 Dataset](https://grouplens.org/datasets/hetrec-2011/) * [Jest Joker Dataset](https://grouplens.org/datasets/jester/) * [Large Movie Review Dataset](http://ai.stanford.edu/~amaas/data/sentiment/) * [MovieLens Dataset](https://grouplens.org/datasets/movielens/) * [Newsgroups Dataset](http://qwone.com/~jason/20Newsgroups/) * [Stanford Large Network Dataset](http://snap.stanford.edu/data/) * [Serendipity 2018 Dataset](https://grouplens.org/datasets/serendipity-2018/) * [Wikilens Dataset](https://grouplens.org/datasets/wikilens/) * [Yelp Dataset](https://www.yelp.com/dataset) * [Yongfeng Zhang Dataset](http://yongfeng.me/dataset/)