[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"similar-business-science--modeltime":3,"tool-business-science--modeltime":61},[4,18,26,36,44,53],{"id":5,"name":6,"github_repo":7,"description_zh":8,"stars":9,"difficulty_score":10,"last_commit_at":11,"category_tags":12,"status":17},4358,"openclaw","openclaw\u002Fopenclaw","OpenClaw 是一款专为个人打造的本地化 AI 助手，旨在让你在自己的设备上拥有完全可控的智能伙伴。它打破了传统 AI 助手局限于特定网页或应用的束缚，能够直接接入你日常使用的各类通讯渠道，包括微信、WhatsApp、Telegram、Discord、iMessage 等数十种平台。无论你在哪个聊天软件中发送消息，OpenClaw 都能即时响应，甚至支持在 macOS、iOS 和 Android 设备上进行语音交互，并提供实时的画布渲染功能供你操控。\n\n这款工具主要解决了用户对数据隐私、响应速度以及“始终在线”体验的需求。通过将 AI 部署在本地，用户无需依赖云端服务即可享受快速、私密的智能辅助，真正实现了“你的数据，你做主”。其独特的技术亮点在于强大的网关架构，将控制平面与核心助手分离，确保跨平台通信的流畅性与扩展性。\n\nOpenClaw 非常适合希望构建个性化工作流的技术爱好者、开发者，以及注重隐私保护且不愿被单一生态绑定的普通用户。只要具备基础的终端操作能力（支持 macOS、Linux 及 Windows WSL2），即可通过简单的命令行引导完成部署。如果你渴望拥有一个懂你",349277,3,"2026-04-06T06:32:30",[13,14,15,16],"Agent","开发框架","图像","数据工具","ready",{"id":19,"name":20,"github_repo":21,"description_zh":22,"stars":23,"difficulty_score":10,"last_commit_at":24,"category_tags":25,"status":17},3808,"stable-diffusion-webui","AUTOMATIC1111\u002Fstable-diffusion-webui","stable-diffusion-webui 是一个基于 Gradio 构建的网页版操作界面，旨在让用户能够轻松地在本地运行和使用强大的 Stable Diffusion 图像生成模型。它解决了原始模型依赖命令行、操作门槛高且功能分散的痛点，将复杂的 AI 绘图流程整合进一个直观易用的图形化平台。\n\n无论是希望快速上手的普通创作者、需要精细控制画面细节的设计师，还是想要深入探索模型潜力的开发者与研究人员，都能从中获益。其核心亮点在于极高的功能丰富度：不仅支持文生图、图生图、局部重绘（Inpainting）和外绘（Outpainting）等基础模式，还独创了注意力机制调整、提示词矩阵、负向提示词以及“高清修复”等高级功能。此外，它内置了 GFPGAN 和 CodeFormer 等人脸修复工具，支持多种神经网络放大算法，并允许用户通过插件系统无限扩展能力。即使是显存有限的设备，stable-diffusion-webui 也提供了相应的优化选项，让高质量的 AI 艺术创作变得触手可及。",162132,"2026-04-05T11:01:52",[14,15,13],{"id":27,"name":28,"github_repo":29,"description_zh":30,"stars":31,"difficulty_score":32,"last_commit_at":33,"category_tags":34,"status":17},1381,"everything-claude-code","affaan-m\u002Feverything-claude-code","everything-claude-code 是一套专为 AI 编程助手（如 Claude Code、Codex、Cursor 等）打造的高性能优化系统。它不仅仅是一组配置文件，而是一个经过长期实战打磨的完整框架，旨在解决 AI 代理在实际开发中面临的效率低下、记忆丢失、安全隐患及缺乏持续学习能力等核心痛点。\n\n通过引入技能模块化、直觉增强、记忆持久化机制以及内置的安全扫描功能，everything-claude-code 能显著提升 AI 在复杂任务中的表现，帮助开发者构建更稳定、更智能的生产级 AI 代理。其独特的“研究优先”开发理念和针对 Token 消耗的优化策略，使得模型响应更快、成本更低，同时有效防御潜在的攻击向量。\n\n这套工具特别适合软件开发者、AI 研究人员以及希望深度定制 AI 工作流的技术团队使用。无论您是在构建大型代码库，还是需要 AI 协助进行安全审计与自动化测试，everything-claude-code 都能提供强大的底层支持。作为一个曾荣获 Anthropic 黑客大奖的开源项目，它融合了多语言支持与丰富的实战钩子（hooks），让 AI 真正成长为懂上",150720,2,"2026-04-11T11:33:10",[14,13,35],"语言模型",{"id":37,"name":38,"github_repo":39,"description_zh":40,"stars":41,"difficulty_score":32,"last_commit_at":42,"category_tags":43,"status":17},2271,"ComfyUI","Comfy-Org\u002FComfyUI","ComfyUI 是一款功能强大且高度模块化的视觉 AI 引擎，专为设计和执行复杂的 Stable Diffusion 图像生成流程而打造。它摒弃了传统的代码编写模式，采用直观的节点式流程图界面，让用户通过连接不同的功能模块即可构建个性化的生成管线。\n\n这一设计巧妙解决了高级 AI 绘图工作流配置复杂、灵活性不足的痛点。用户无需具备编程背景，也能自由组合模型、调整参数并实时预览效果，轻松实现从基础文生图到多步骤高清修复等各类复杂任务。ComfyUI 拥有极佳的兼容性，不仅支持 Windows、macOS 和 Linux 全平台，还广泛适配 NVIDIA、AMD、Intel 及苹果 Silicon 等多种硬件架构，并率先支持 SDXL、Flux、SD3 等前沿模型。\n\n无论是希望深入探索算法潜力的研究人员和开发者，还是追求极致创作自由度的设计师与资深 AI 绘画爱好者，ComfyUI 都能提供强大的支持。其独特的模块化架构允许社区不断扩展新功能，使其成为当前最灵活、生态最丰富的开源扩散模型工具之一，帮助用户将创意高效转化为现实。",108322,"2026-04-10T11:39:34",[14,15,13],{"id":45,"name":46,"github_repo":47,"description_zh":48,"stars":49,"difficulty_score":32,"last_commit_at":50,"category_tags":51,"status":17},6121,"gemini-cli","google-gemini\u002Fgemini-cli","gemini-cli 是一款由谷歌推出的开源 AI 命令行工具，它将强大的 Gemini 大模型能力直接集成到用户的终端环境中。对于习惯在命令行工作的开发者而言，它提供了一条从输入提示词到获取模型响应的最短路径，无需切换窗口即可享受智能辅助。\n\n这款工具主要解决了开发过程中频繁上下文切换的痛点，让用户能在熟悉的终端界面内直接完成代码理解、生成、调试以及自动化运维任务。无论是查询大型代码库、根据草图生成应用，还是执行复杂的 Git 操作，gemini-cli 都能通过自然语言指令高效处理。\n\n它特别适合广大软件工程师、DevOps 人员及技术研究人员使用。其核心亮点包括支持高达 100 万 token 的超长上下文窗口，具备出色的逻辑推理能力；内置 Google 搜索、文件操作及 Shell 命令执行等实用工具；更独特的是，它支持 MCP（模型上下文协议），允许用户灵活扩展自定义集成，连接如图像生成等外部能力。此外，个人谷歌账号即可享受免费的额度支持，且项目基于 Apache 2.0 协议完全开源，是提升终端工作效率的理想助手。",100752,"2026-04-10T01:20:03",[52,13,15,14],"插件",{"id":54,"name":55,"github_repo":56,"description_zh":57,"stars":58,"difficulty_score":32,"last_commit_at":59,"category_tags":60,"status":17},4721,"markitdown","microsoft\u002Fmarkitdown","MarkItDown 是一款由微软 AutoGen 团队打造的轻量级 Python 工具，专为将各类文件高效转换为 Markdown 格式而设计。它支持 PDF、Word、Excel、PPT、图片（含 OCR）、音频（含语音转录）、HTML 乃至 YouTube 链接等多种格式的解析，能够精准提取文档中的标题、列表、表格和链接等关键结构信息。\n\n在人工智能应用日益普及的今天，大语言模型（LLM）虽擅长处理文本，却难以直接读取复杂的二进制办公文档。MarkItDown 恰好解决了这一痛点，它将非结构化或半结构化的文件转化为模型“原生理解”且 Token 效率极高的 Markdown 格式，成为连接本地文件与 AI 分析 pipeline 的理想桥梁。此外，它还提供了 MCP（模型上下文协议）服务器，可无缝集成到 Claude Desktop 等 LLM 应用中。\n\n这款工具特别适合开发者、数据科学家及 AI 研究人员使用，尤其是那些需要构建文档检索增强生成（RAG）系统、进行批量文本分析或希望让 AI 助手直接“阅读”本地文件的用户。虽然生成的内容也具备一定可读性，但其核心优势在于为机器",93400,"2026-04-06T19:52:38",[52,14],{"id":62,"github_repo":63,"name":64,"description_en":65,"description_zh":66,"ai_summary_zh":66,"readme_en":67,"readme_zh":68,"quickstart_zh":69,"use_case_zh":70,"hero_image_url":71,"owner_login":72,"owner_name":73,"owner_avatar_url":74,"owner_bio":75,"owner_company":76,"owner_location":76,"owner_email":77,"owner_twitter":76,"owner_website":78,"owner_url":79,"languages":80,"stars":89,"forks":90,"last_commit_at":91,"license":92,"difficulty_score":93,"env_os":94,"env_gpu":95,"env_ram":95,"env_deps":96,"category_tags":108,"github_topics":110,"view_count":32,"oss_zip_url":76,"oss_zip_packed_at":76,"status":17,"created_at":126,"updated_at":127,"faqs":128,"releases":154},6633,"business-science\u002Fmodeltime","modeltime","Modeltime unlocks time series forecast models and machine learning in one framework","Modeltime 是一个专为 R 语言打造的时间序列预测框架，旨在将传统统计模型与机器学习算法统一在一个简洁高效的工作流中。它解决了以往用户需要在不同框架间切换、难以同时利用经典时间序列方法（如 ARIMA、ETS）和现代机器学习模型（如随机森林、梯度提升树、Facebook Prophet）的痛点。\n\n通过 Modeltime，数据分析师、研究人员和开发者可以遵循标准化的六步流程，轻松完成从数据准备、模型训练到预测评估的全过程。其核心亮点在于深度集成 tidymodels 生态系统，允许用户直接调用 parsnip 支持的各类机器学习模型进行时序预测，同时保持代码风格的一致性与可维护性。此外，Modeltime 还拥有一个不断扩展的工具生态，支持自动机器学习（H2O）、深度学习（GluonTS）以及模型融合等高级功能，满足从基础分析到复杂建模的多样化需求。\n\n无论你是希望快速上手时间序列预测的初学者，还是追求高性能与可扩展性的专业数据科学家，Modeltime 都能提供友好而强大的支持，让高质量的时间序列分析变得更简单、更快速。","\n\u003C!-- README.md is generated from README.Rmd. Please edit that file -->\n\n# modeltime\n\n\u003C!-- badges: start -->\n\n[![CRAN_Status_Badge](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fbusiness-science_modeltime_readme_2ee118e35eca.png)](https:\u002F\u002Fcran.r-project.org\u002Fpackage=modeltime)\n![](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fbusiness-science_modeltime_readme_8e30c8535d26.png)\n![](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fbusiness-science_modeltime_readme_67324a5151a0.png)\n[![Codecov test\ncoverage](https:\u002F\u002Fcodecov.io\u002Fgh\u002Fbusiness-science\u002Fmodeltime\u002Fbranch\u002Fmaster\u002Fgraph\u002Fbadge.svg)](https:\u002F\u002Fapp.codecov.io\u002Fgh\u002Fbusiness-science\u002Fmodeltime?branch=master)\n[![R-CMD-check](https:\u002F\u002Fgithub.com\u002Fbusiness-science\u002Fmodeltime\u002Fworkflows\u002FR-CMD-check\u002Fbadge.svg)](https:\u002F\u002Fgithub.com\u002Fbusiness-science\u002Fmodeltime\u002Factions)\n\u003C!-- badges: end -->\n\n> Tidy time series forecasting in `R`.\n\nMission: Our number 1 goal is to make high-performance time series\nanalysis easier, faster, and more scalable. Modeltime solves this with a\nsimple to use infrastructure for modeling and forecasting time series.\n\n## Quickstart Video\n\nFor those that prefer video tutorials, we have an [11-minute YouTube\nVideo](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=-bCelif-ENY) that walks you\nthrough the Modeltime Workflow.\n\n\u003Ca href=\"https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=-bCelif-ENY\" target=\"_blank\">\n\u003Cp style=\"text-align:center;\">\n\u003Cimg src= \"vignettes\u002Fmodeltime-video.jpg\"\nalt=\"Introduction to Modeltime\" width=\"60%\"\u002F>\n\u003C\u002Fp>\n\u003Cp style=\"text-align:center\">\n(Click to Watch on YouTube)\n\u003C\u002Fp>\n\n\u003C\u002Fa>\n\n## Tutorials\n\n- [**Getting Started with\n  Modeltime**](https:\u002F\u002Fbusiness-science.github.io\u002Fmodeltime\u002Farticles\u002Fgetting-started-with-modeltime.html):\n  A walkthrough of the 6-Step Process for using `modeltime` to forecast\n\n- [**Modeltime\n  Documentation**](https:\u002F\u002Fbusiness-science.github.io\u002Fmodeltime\u002F): Learn\n  how to **use** `modeltime`, **find** *Modeltime Models*, and\n  **extend** `modeltime` so you can use new algorithms inside the\n  *Modeltime Workflow*.\n\n## Installation\n\nCRAN version:\n\n``` r\ninstall.packages(\"modeltime\", dependencies = TRUE)\n```\n\nDevelopment version:\n\n``` r\nremotes::install_github(\"business-science\u002Fmodeltime\", dependencies = TRUE)\n```\n\n## Why modeltime?\n\n> Modeltime unlocks time series models and machine learning in one\n> framework\n\n\u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fbusiness-science_modeltime_readme_f2c4474e5e55.jpg\" width=\"100%\" style=\"display: block; margin: auto;\" \u002F>\n\nNo need to switch back and forth between various frameworks. `modeltime`\nunlocks machine learning & classical time series analysis.\n\n- **forecast**: Use ARIMA, ETS, and more models coming (`arima_reg()`,\n  `arima_boost()`, & `exp_smoothing()`).\n- **prophet**: Use Facebook’s Prophet algorithm (`prophet_reg()` &\n  `prophet_boost()`)\n- **tidymodels**: Use any `parsnip` model: `rand_forest()`,\n  `boost_tree()`, `linear_reg()`, `mars()`, `svm_rbf()` to forecast\n\n## Forecast faster\n\n> A streamlined workflow for forecasting\n\nModeltime incorporates a [streamlined workflow (see Getting Started with\nModeltime)](https:\u002F\u002Fbusiness-science.github.io\u002Fmodeltime\u002Farticles\u002Fgetting-started-with-modeltime.html)\nfor using best practices to forecast.\n\n\u003Chr>\n\n\u003Cdiv class=\"figure\" style=\"text-align: center\">\n\n\u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fbusiness-science_modeltime_readme_8359fa45bdee.jpg\" alt=\"A streamlined workflow for forecasting\" width=\"100%\" \u002F>\n\u003Cp class=\"caption\">\nA streamlined workflow for forecasting\n\u003C\u002Fp>\n\n\u003C\u002Fdiv>\n\n\u003Chr>\n\n## Meet the modeltime ecosystem\n\n> Learn a growing ecosystem of forecasting packages\n\n\u003Cdiv class=\"figure\" style=\"text-align: center\">\n\n\u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fbusiness-science_modeltime_readme_2273c4994b6d.jpg\" alt=\"The modeltime ecosystem is growing\" width=\"100%\" \u002F>\n\u003Cp class=\"caption\">\nThe modeltime ecosystem is growing\n\u003C\u002Fp>\n\n\u003C\u002Fdiv>\n\nModeltime is part of a **growing ecosystem** of Modeltime forecasting\npackages.\n\n- [Modeltime (Machine\n  Learning)](https:\u002F\u002Fbusiness-science.github.io\u002Fmodeltime\u002F)\n\n- [Modeltime H2O\n  (AutoML)](https:\u002F\u002Fbusiness-science.github.io\u002Fmodeltime.h2o\u002F)\n\n- [Modeltime GluonTS (Deep\n  Learning)](https:\u002F\u002Fbusiness-science.github.io\u002Fmodeltime.gluonts\u002F)\n\n- [Modeltime Ensemble (Blending\n  Forecasts)](https:\u002F\u002Fbusiness-science.github.io\u002Fmodeltime.ensemble\u002F)\n\n- [Modeltime Resample\n  (Backtesting)](https:\u002F\u002Fbusiness-science.github.io\u002Fmodeltime.resample\u002F)\n\n- [Timetk (Feature Engineering, Data Wrangling, Time Series\n  Visualization)](https:\u002F\u002Fbusiness-science.github.io\u002Ftimetk\u002F)\n\n## Summary\n\nModeltime is an amazing ecosystem for time series forecasting. But it\ncan take a long time to learn:\n\n- Many algorithms\n- Ensembling and Resampling\n- Machine Learning\n- Deep Learning\n- Scalable Modeling: 10,000+ time series\n\nYour probably thinking how am I ever going to learn time series\nforecasting. Here’s the solution that will save you years of struggling.\n\n## Take the High-Performance Forecasting Course\n\n> Become the forecasting expert for your organization\n\n\u003Ca href=\"https:\u002F\u002Funiversity.business-science.io\u002Fp\u002Fds4b-203-r-high-performance-time-series-forecasting\u002F\" target=\"_blank\">\u003Cimg src=\"https:\u002F\u002Fwww.filepicker.io\u002Fapi\u002Ffile\u002FbKyqVAi5Qi64sS05QYLk\" alt=\"High-Performance Time Series Forecasting Course\" width=\"100%\" style=\"box-shadow: 0 0 5px 2px rgba(0, 0, 0, .5);\"\u002F>\u003C\u002Fa>\n\n[*High-Performance Time Series\nCourse*](https:\u002F\u002Funiversity.business-science.io\u002Fp\u002Fds4b-203-r-high-performance-time-series-forecasting\u002F)\n\n### Time Series is Changing\n\nTime series is changing. **Businesses now need 10,000+ time series\nforecasts every day.** This is what I call a *High-Performance Time\nSeries Forecasting System (HPTSF)* - Accurate, Robust, and Scalable\nForecasting.\n\n**High-Performance Forecasting Systems will save companies by improving\naccuracy and scalability.** Imagine what will happen to your career if\nyou can provide your organization a “High-Performance Time Series\nForecasting System” (HPTSF System).\n\n### How to Learn High-Performance Time Series Forecasting\n\nI teach how to build a HPTFS System in my [**High-Performance Time\nSeries Forecasting\nCourse**](https:\u002F\u002Funiversity.business-science.io\u002Fp\u002Fds4b-203-r-high-performance-time-series-forecasting).\nYou will learn:\n\n- **Time Series Machine Learning** (cutting-edge) with `Modeltime` - 30+\n  Models (Prophet, ARIMA, XGBoost, Random Forest, & many more)\n- **Deep Learning** with `GluonTS` (Competition Winners)\n- **Time Series Preprocessing**, Noise Reduction, & Anomaly Detection\n- **Feature engineering** using lagged variables & external regressors\n- **Hyperparameter Tuning**\n- **Time series cross-validation**\n- **Ensembling** Multiple Machine Learning & Univariate Modeling\n  Techniques (Competition Winner)\n- **Scalable Forecasting** - Forecast 1000+ time series in parallel\n- and more.\n\n\u003Cp class=\"text-center\" style=\"font-size:24px;\">\nBecome the Time Series Expert for your organization.\n\u003C\u002Fp>\n\u003Cbr>\n\u003Cp class=\"text-center\" style=\"font-size:30px;\">\n\u003Ca href=\"https:\u002F\u002Funiversity.business-science.io\u002Fp\u002Fds4b-203-r-high-performance-time-series-forecasting\">Take\nthe High-Performance Time Series Forecasting Course\u003C\u002Fa>\n\u003C\u002Fp>\n","\u003C!-- README.md 由 README.Rmd 生成。请编辑该文件 -->\n\n# modeltime\n\n\u003C!-- badges: start -->\n\n[![CRAN_Status_Badge](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fbusiness-science_modeltime_readme_2ee118e35eca.png)](https:\u002F\u002Fcran.r-project.org\u002Fpackage=modeltime)\n![](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fbusiness-science_modeltime_readme_8e30c8535d26.png)\n![](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fbusiness-science_modeltime_readme_67324a5151a0.png)\n[![Codecov 测试\n覆盖率](https:\u002F\u002Fcodecov.io\u002Fgh\u002Fbusiness-science\u002Fmodeltime\u002Fbranch\u002Fmaster\u002Fgraph\u002Fbadge.svg)](https:\u002F\u002Fapp.codecov.io\u002Fgh\u002Fbusiness-science\u002Fmodeltime?branch=master)\n[![R-CMD-check](https:\u002F\u002Fgithub.com\u002Fbusiness-science\u002Fmodeltime\u002Fworkflows\u002FR-CMD-check\u002Fbadge.svg)](https:\u002F\u002Fgithub.com\u002Fbusiness-science\u002Fmodeltime\u002Factions)\n\u003C!-- badges: end -->\n\n> 在 `R` 中进行整洁的时间序列预测。\n\n使命：我们的首要目标是让高性能的时间序列分析更加简单、快速且可扩展。Modeltime 通过一套易于使用的基础设施来建模和预测时间序列，从而实现这一目标。\n\n## 快速入门视频\n\n对于喜欢视频教程的用户，我们提供了一个 [11 分钟的 YouTube 视频](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=-bCelif-ENY)，带您逐步了解 Modeltime 工作流程。\n\n\u003Ca href=\"https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=-bCelif-ENY\" target=\"_blank\">\n\u003Cp style=\"text-align:center;\">\n\u003Cimg src= \"vignettes\u002Fmodeltime-video.jpg\"\nalt=\"Modeltime 简介\" width=\"60%\"\u002F>\n\u003C\u002Fp>\n\u003Cp style=\"text-align:center\">\n（点击观看 YouTube）\n\u003C\u002Fp>\n\n\u003C\u002Fa>\n\n## 教程\n\n- [**Modeltime 入门**](https:\u002F\u002Fbusiness-science.github.io\u002Fmodeltime\u002Farticles\u002Fgetting-started-with-modeltime.html)：\n  介绍使用 `modeltime` 进行预测的 6 步流程。\n\n- [**Modeltime 文档**](https:\u002F\u002Fbusiness-science.github.io\u002Fmodeltime\u002F)：\n  学习如何 **使用** `modeltime`、**查找** *Modeltime 模型*，以及 **扩展** `modeltime`，以便在 *Modeltime 工作流程* 中使用新的算法。\n\n## 安装\n\nCRAN 版本：\n\n``` r\ninstall.packages(\"modeltime\", dependencies = TRUE)\n```\n\n开发版本：\n\n``` r\nremotes::install_github(\"business-science\u002Fmodeltime\", dependencies = TRUE)\n```\n\n## 为什么选择 modeltime？\n\n> Modeltime 将时间序列模型和机器学习整合到一个框架中\n\n\u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fbusiness-science_modeltime_readme_f2c4474e5e55.jpg\" width=\"100%\" style=\"display: block; margin: auto;\" \u002F>\n\n无需在不同的框架之间来回切换。`modeltime` 让机器学习和经典的时间序列分析得以统一。\n\n- **forecast**：使用 ARIMA、ETS 等模型，未来还将支持更多模型（`arima_reg()`、`arima_boost()` 和 `exp_smoothing()`）。\n- **prophet**：使用 Facebook 的 Prophet 算法（`prophet_reg()` 和 `prophet_boost()`）。\n- **tidymodels**：使用任何 `parsnip` 模型：`rand_forest()`、`boost_tree()`、`linear_reg()`、`mars()`、`svm_rbf()` 等进行预测。\n\n## 更快地进行预测\n\n> 用于预测的简化工作流程\n\nModeltime 提供了一个[简化的预测工作流程（参见 Modeltime 入门）](https:\u002F\u002Fbusiness-science.github.io\u002Fmodeltime\u002Farticles\u002Fgetting-started-with-modeltime.html)，帮助您采用最佳实践进行预测。\n\n\u003Chr>\n\n\u003Cdiv class=\"figure\" style=\"text-align: center\">\n\n\u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fbusiness-science_modeltime_readme_8359fa45bdee.jpg\" alt=\"简化的预测工作流程\" width=\"100%\" \u002F>\n\u003Cp class=\"caption\">\n简化的预测工作流程\n\u003C\u002Fp>\n\n\u003C\u002Fdiv>\n\n\u003Chr>\n\n## 认识 Modeltime 生态系统\n\n> 了解不断发展的预测工具生态系统\n\n\u003Cdiv class=\"figure\" style=\"text-align: center\">\n\n\u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fbusiness-science_modeltime_readme_2273c4994b6d.jpg\" alt=\"Modeltime 生态系统正在发展\" width=\"100%\" \u002F>\n\u003Cp class=\"caption\">\nModeltime 生态系统正在发展\n\u003C\u002Fp>\n\n\u003C\u002Fdiv>\n\nModeltime 是一个不断增长的 **Modeltime 预测工具生态系统**的一部分。\n\n- [Modeltime（机器学习）](https:\u002F\u002Fbusiness-science.github.io\u002Fmodeltime\u002F)\n\n- [Modeltime H2O（AutoML）](https:\u002F\u002Fbusiness-science.github.io\u002Fmodeltime.h2o\u002F)\n\n- [Modeltime GluonTS（深度学习）](https:\u002F\u002Fbusiness-science.github.io\u002Fmodeltime.gluonts\u002F)\n\n- [Modeltime Ensemble（预测融合）](https:\u002F\u002Fbusiness-science.github.io\u002Fmodeltime.ensemble\u002F)\n\n- [Modeltime Resample（回测）](https:\u002F\u002Fbusiness-science.github.io\u002Fmodeltime.resample\u002F)\n\n- [Timetk（特征工程、数据整理、时间序列可视化）](https:\u002F\u002Fbusiness-science.github.io\u002Ftimetk\u002F)\n\n## 总结\n\nModeltime 是一个强大的时间序列预测生态系统。然而，掌握它可能需要较长时间：\n\n- 涉及多种算法\n- 预测融合与重采样\n- 机器学习\n- 深度学习\n- 可扩展建模：处理上万条时间序列\n\n您可能会想，我该如何学会时间序列预测呢？这里有一个能为您节省多年摸索时间的解决方案。\n\n## 参加高性能预测课程\n\n> 成为贵组织的预测专家\n\n\u003Ca href=\"https:\u002F\u002Funiversity.business-science.io\u002Fp\u002Fds4b-203-r-high-performance-time-series-forecasting\u002F\" target=\"_blank\">\u003Cimg src=\"https:\u002F\u002Fwww.filepicker.io\u002Fapi\u002Ffile\u002FbKyqVAi5Qi64sS05QYLk\" alt=\"高性能时间序列预测课程\" width=\"100%\" style=\"box-shadow: 0 0 5px 2px rgba(0, 0, 0, .5);\"\u002F>\u003C\u002Fa>\n\n[*高性能时间序列预测课程*](https:\u002F\u002Funiversity.business-science.io\u002Fp\u002Fds4b-203-r-high-performance-time-series-forecasting\u002F)\n\n### 时间序列正在变革\n\n时间序列正在发生变化。**如今，企业每天需要进行超过一万次的时间序列预测。** 我称之为“高性能时间序列预测系统”（HPTSF）——准确、稳健且可扩展的预测。\n\n**高性能预测系统将通过提高准确性和可扩展性为企业节省成本。** 想象一下，如果您能够为您的组织提供一个“高性能时间序列预测系统”（HPTSF 系统），您的职业生涯将会发生怎样的变化。\n\n### 如何学习高性能时间序列预测\n\n我在我的[**高性能时间序列预测课程**](https:\u002F\u002Funiversity.business-science.io\u002Fp\u002Fds4b-203-r-high-performance-time-series-forecasting) 中教授如何构建 HPTFS 系统。您将学习：\n\n- 使用 `Modeltime` 的 **时间序列机器学习**（最前沿技术）——包含 30 多种模型（Prophet、ARIMA、XGBoost、随机森林等）。\n- 使用 `GluonTS` 的 **深度学习**（竞赛获奖技术）。\n- **时间序列预处理**、降噪和异常检测。\n- 使用滞后变量和外部回归因子进行 **特征工程**。\n- **超参数调优**。\n- **时间序列交叉验证**。\n- 融合多种机器学习和单变量建模技术（竞赛获奖方法）。\n- **可扩展预测**——并行预测上千条时间序列。\n- 以及其他内容。\n\n\u003Cp class=\"text-center\" style=\"font-size:24px;\">\n成为贵组织的时间序列专家。\n\u003C\u002Fp>\n\u003Cbr>\n\u003Cp class=\"text-center\" style=\"font-size:30px;\">\n\u003Ca href=\"https:\u002F\u002Funiversity.business-science.io\u002Fp\u002Fds4b-203-r-high-performance-time-series-forecasting\">立即参加高性能时间序列预测课程\u003C\u002Fa>\n\u003C\u002Fp>","# Modeltime 快速上手指南\n\nModeltime 是一个用于 R 语言的高效时间序列预测框架，旨在简化机器学习与经典时间序列模型（如 ARIMA、Prophet）的统一调用流程，支持大规模并行预测。\n\n## 环境准备\n\n- **操作系统**：Windows、macOS 或 Linux\n- **R 版本**：建议 R 4.0 及以上版本\n- **前置依赖**：\n  - `tidymodels` 生态系统（含 `parsnip`, `recipes`, `workflows` 等）\n  - `timetk`（用于时间序列特征工程与可视化）\n  - 可选：`prophet`、`xgboost`、`randomForest` 等具体算法包（Modeltime 会自动处理大部分依赖）\n\n> 💡 提示：国内用户可配置 CRAN 镜像加速安装，例如使用清华大学镜像：\n> ```r\n> options(repos = c(CRAN = \"https:\u002F\u002Fmirrors.tuna.tsinghua.edu.cn\u002FCRAN\"))\n> ```\n\n## 安装步骤\n\n### 方式一：安装 CRAN 稳定版（推荐）\n\n```r\ninstall.packages(\"modeltime\", dependencies = TRUE)\n```\n\n### 方式二：安装 GitHub 开发版（获取最新功能）\n\n```r\nremotes::install_github(\"business-science\u002Fmodeltime\", dependencies = TRUE)\n```\n\n> 若未安装 `remotes` 包，请先运行：`install.packages(\"remotes\")`\n\n## 基本使用\n\n以下是最简化的 6 步预测流程示例（基于内置数据集 `m750`）：\n\n```r\nlibrary(modeltime)\nlibrary(timetk)\nlibrary(tidymodels)\n\n# 1. 加载数据\ndata(m750)\n\n# 2. 划分训练集与测试集\nsplit \u003C- time_series_split(m750, assess = \"3 months\", cumulative = TRUE)\n\n# 3. 定义模型（以随机森林为例）\nwf_spec \u003C- workflow() %>%\n  add_model(\n    rand_forest(mode = \"regression\") %>% set_engine(\"ranger\")\n  ) %>%\n  add_recipe(\n    recipe(value ~ date, data = m750) %>%\n      step_date(date, features = c(\"month\", \"dow\")) %>%\n      step_lag(value, lag = 1:12)\n  )\n\n# 4. 创建 Modeltime Table\nmodel_table \u003C- modeltime_table(\n  wf_spec\n)\n\n# 5. 拟合模型\nfitted_table \u003C- model_table %>%\n  modeltime_fit_resamples(\n    resamples = split,\n    control = control_resamples(save_pred = TRUE)\n  )\n\n# 6. 生成预测\nforecast_results \u003C- fitted_table %>%\n  modeltime_forecast(\n    new_data = m750,\n    actual_data = m750\n  )\n\n# 可视化结果\nforecast_results %>%\n  plot_modeltime_forecast(.facet_by = \"id\", .interactive = FALSE)\n```\n\n该流程统一了数据预处理、模型训练、交叉验证与预测输出，支持轻松切换至 ARIMA、Prophet、XGBoost 等数十种算法，并可扩展至成千上万条时间序列的并行预测。","某零售连锁企业的数据团队需要为旗下 500 家门店预测未来三个月的每日销量，以优化库存管理和采购计划。\n\n### 没有 modeltime 时\n- **框架割裂严重**：尝试经典统计模型（如 ARIMA）需使用 `forecast` 包，而想引入机器学习（如随机森林）则需切换至 `tidymodels` 或手动编写代码，导致工作流支离破碎。\n- **多序列处理繁琐**：面对 500 家门店的海量时间序列数据，缺乏统一的批量建模接口，只能编写复杂的循环嵌套代码，极易出错且运行缓慢。\n- **模型对比困难**：不同算法输出的结果格式各异，难以在同一框架下快速评估和对比哪种模型更适合特定门店的销售模式。\n- **扩展性差**：若想尝试 Facebook Prophet 或深度学习模型，往往需要重新搭建整套环境，无法在现有流程中无缝集成。\n\n### 使用 modeltime 后\n- **统一建模框架**：modeltime 将 ARIMA、Prophet 及各类机器学习算法（通过 `parsnip`）整合在同一接口下，无需切换工具即可自由调用多种策略。\n- **高效批量预测**：利用其标准化的六步工作流，可轻松对 500 家门店的时间序列进行分组并行建模，大幅缩短从数据准备到输出预测值的时间。\n- **标准化评估体系**：所有模型的预测结果自动转换为统一的 Tidy 格式，便于直接使用可视化工具对比误差指标，快速锁定最优方案。\n- **生态无缝扩展**：通过 modeltime 生态系统（如 modeltime.h2o 或 modeltime.gluonts），能平滑接入 AutoML 或深度学习能力，无需重构现有代码逻辑。\n\nmodeltime 通过统一的时间序列分析框架，让数据团队能以更低成本、更高效率实现从传统统计到前沿机器学习的规模化落地。","https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fbusiness-science_modeltime_f2c4474e.jpg","business-science","Business Science","https:\u002F\u002Foss.gittoolsai.com\u002Favatars\u002Fbusiness-science_91afe14c.png","Applying data science to business & financial analysis, tw: @bizScienc",null,"info@business-science.io","www.business-science.io","https:\u002F\u002Fgithub.com\u002Fbusiness-science",[81,85],{"name":82,"color":83,"percentage":84},"R","#198CE7",99.1,{"name":86,"color":87,"percentage":88},"CSS","#663399",0.9,576,86,"2026-04-10T02:48:41","NOASSERTION",1,"Linux, macOS, Windows","未说明",{"notes":97,"python":98,"dependencies":99},"该工具是基于 R 语言的包，非 Python 项目。安装需通过 CRAN 或 GitHub。虽然核心包支持经典时间序列和机器学习模型，但若需使用深度学习功能（如 GluonTS）或自动机器学习（H2O），需额外安装对应的扩展包，这些扩展包可能涉及特定的后端依赖。官方推荐参加相关课程以掌握处理 10,000+ 时间序列的高性能预测系统构建方法。","不适用 (基于 R 语言)",[100,101,102,103,104,105,106,107],"R (核心环境)","tidymodels","parsnip","forecast","prophet","timetk","modeltime.h2o (可选)","modeltime.gluonts (可选)",[35,14,16,109],"其他",[111,112,113,101,114,115,104,116,117,118,64,119,120,121,122,123,124,125],"time","time-series","forecasting","machine-learning-algorithms","machine-learning","arima","ets","tbats","tidymodeling","data-science","deep-learning","time-series-analysis","timeseries","timeseries-forecasting","r-package","2026-03-27T02:49:30.150509","2026-04-11T23:23:56.196841",[129,134,139,144,149],{"id":130,"question_zh":131,"answer_zh":132,"source_url":133},29944,"为什么在使用 `modeltime_calibrate()` 校准线性回归模型时出现 `.pred` 对象未找到的错误？","这是因为 tidyverse 的 predict 函数最新版本返回的是 `.pred_res` 列而不是 `.pred` 列，导致 modeltime 无法识别。解决方案是更新 modeltime 包到最新版本，维护者已发布新版本以兼容此变化，直到 tidymodels 团队修复 `.pred_res` 问题前均可正常使用。请运行 `install.packages(\"modeltime\")` 更新。","https:\u002F\u002Fgithub.com\u002Fbusiness-science\u002Fmodeltime\u002Fissues\u002F228",{"id":135,"question_zh":136,"answer_zh":137,"source_url":138},29945,"如何在 modeltime 中正确使用并行重拟合（refit）功能，避免只使用单核的问题？","在使用 `parallel_start()` 启动并行计算时，可以通过传递参数指定集群类型来解决多核未充分利用的问题。例如，使用 `parallel_start(type=\"FORK\")`。该参数会直接传递给底层的 `makeCluster()` 函数。确保在 `control_refit()` 中设置 `allow_par = TRUE` 并指定 `cores` 数量。","https:\u002F\u002Fgithub.com\u002Fbusiness-science\u002Fmodeltime\u002Fissues\u002F113",{"id":140,"question_zh":141,"answer_zh":142,"source_url":143},29946,"加载 modeltime 或 tidymodels 时出现命名空间版本冲突错误（如 vctrs, rlang, dplyr 版本过低）怎么办？","这通常是由于本地安装的依赖包版本过旧导致的，并非 modeltime 本身的 bug。解决方法是手动更新所有相关的依赖包。可以运行 `update.packages()` 或单独安装所需版本，例如确保 `vctr >= 0.3.0`, `rlang >= 0.4.7`, `dplyr >= 1.0.0`。更新后重启 R 会话再次加载包即可。","https:\u002F\u002Fgithub.com\u002Fbusiness-science\u002Fmodeltime\u002Fissues\u002F35",{"id":145,"question_zh":146,"answer_zh":147,"source_url":148},29947,"modeltime 是否支持广义加性模型（GAMs\u002Fmgcv）？","是的，目前 GAM 模型的支持已经通过 `parsnip` 包实现集成。用户可以直接使用 parsnip 定义的 GAM 引擎配合 modeltime 工作流进行建模。对于拟合好的 gam 模型对象，可以像 lm 模型一样使用 `summary()`, `plot()`, `predict()` 函数，同时也支持 `broom` 包的 `tidy()`, `glance()`, `augment()` 函数进行处理。","https:\u002F\u002Fgithub.com\u002Fbusiness-science\u002Fmodeltime\u002Fissues\u002F71",{"id":150,"question_zh":151,"answer_zh":152,"source_url":153},29948,"modeltime 是否支持对 ARIMA 等非全局模型进行迭代式预测（Iterative Forecasting）？","支持。Modeltime 已经集成了迭代预测系统，专门用于解决 ARIMA、指数平滑等不适合面板数据（全局模型）的算法扩展问题。该功能允许用户在循环中对每个序列单独训练和预测。此功能已在开发版中完成，并包含在 modeltime 1.0.0 及更高版本中。工作流程包括数据嵌套、迭代拟合、精度评估和未来预测等步骤。","https:\u002F\u002Fgithub.com\u002Fbusiness-science\u002Fmodeltime\u002Fissues\u002F122",[155,160,165,170,175,180,185,190,195,200,205,210,215,220,225,230,235,240],{"id":156,"version":157,"summary_zh":158,"released_at":159},206539,"v1.3.3","## 变更内容\n* 通过 @EmilHvitfeldt 在 https:\u002F\u002Fgithub.com\u002Fbusiness-science\u002Fmodeltime\u002Fpull\u002F263 中的提交，使软件包能够更好地应对 xgboost 版本的变化。\n\n\n**完整变更日志**: https:\u002F\u002Fgithub.com\u002Fbusiness-science\u002Fmodeltime\u002Fcompare\u002Fv1.3.2...v1.3.3","2025-12-18T19:34:39",{"id":161,"version":162,"summary_zh":163,"released_at":164},206540,"v1.3.2","# modeltime 1.3.2\n\n### 亮点\n\n* **用于并行处理的 future 后端。** `parallel_start()` 现在支持 `.method = \"future\"`（通过 `doFuture` 桥接），并且内部辅助函数在可用时会优先使用 `future::multisession` 计划。这减少了 foreach\u002Ffuture 调优警告，并使并行设置更具可移植性。\n* **新增指标：`maape()`。** 新增平均反正切绝对百分比误差，并将其纳入 `extended_forecast_accuracy_metric_set()` 中。完全兼容 yardstick ≥ 1.2.0（支持 `case_weights`），同时保持对旧版 yardstick 的向后兼容性。\n* **ADAM 调参参数。** 为 `adam_reg()` 引擎新增 dials 辅助函数 `ets_model()` 和 `loss()`，以提供更丰富的模型选择和损失函数选项。\n\n### 新功能\n\n* `parallel_start(..., .method = \"future\")`：设置 `future::multisession` 计划，并在存在时注册 `doFuture`。\n* 导出指标：`maape()`、`maape_vec()`。\n* Dials 参数：`ets_model()`、`loss()`。\n\n### 改进\n\n* **与 yardstick 兼容性：** 内部包装器避免了在 yardstick ≥ 1.2.0 下出现的弃用和错误（不再出现“选择中包含外部向量”或缺少 `case_weights` 的问题）。\n\n  * `summarize_accuracy_metrics()` 现在使用 `dplyr::any_of(\".estimator\")`，以确保在不同版本的 yardstick 中都能正常运行。\n* **调参网格：** 测试已从 `dials::grid_latin_hypercube()` 迁移到 `dials::grid_space_filling()`。\n* **并行控制用户体验：** `setup_parallel_processing()` 在可用时优先使用 `future`；`parallel_stop()` 则会优雅地恢复为顺序执行。\n* **文档优化：** Roxygen 链接标准化（例如，函数引用使用反引号）；绘图包装函数在文档中明确引用 `timetk::*`。\n* **实用工具恢复：** 重新添加了 `calc_accuracy_2()`，以修复引用该函数的小插件构建问题。\n\n### 错误修复\n\n* 修复了在 yardstick 1.2+ 下计算 MAAPE 的问题（处理 `case_weights` 的 tidyselect），消除了“必须至少选择一项”的错误提示。\n* 解决了小插件构建失败的问题（未找到 `calc_accuracy_2`）。\n* 通过优先使用 `future` 后端，减少了测试中多余的 foreach RNG 警告；在顺序运行时，消息提示也得到了改进。\n* 稳定了 ADAM 引擎参数的处理方式（对于多选参数，会防御性地选择第一个）。\n\n### 维护与 CI\n\n* **建议包：** 添加了 `future` 和 `doFuture`，并保留了 `TSrepr`。\n* **CI：** 更新了 GitHub Actions 步骤（`actions\u002Fcheckout@v4`、`actions\u002Fupload-artifact@v4`），修复了少量路径问题，并增加了可选的缓存框架。\n* **Roxygen：** 更新至 7.3.2。\n* **数据文档：** 更新了 M4 竞赛的 URL。\n\n### 破坏性变更\n\n* 无。\n\n### 已弃用内容\n\n* 无。\n\n### 迁移说明\n\n如果您使用并行调参或重拟合，请优先使用新的 future 后端：\n\n```r\n# 设置并行处理（跨操作系统通用）\nparallel_start(2, .method = \"future\")  # 或省略 2 以使用所有可用物理核心\n# ... 运行 tune_*()、modeltime_*() 等 ...\nparallel_stop()\n```\n\n`extended_forecast_accuracy_metric_set()` 现默认包含 `maape()`；如果您之前提供了自定义的 MAAPE，可以将其移除。","2025-08-29T11:49:03",{"id":166,"version":167,"summary_zh":168,"released_at":169},206541,"v1.3.0","# modeltime 1.3.0\n\n#### 概述\n\n此版本与 modeltime 1.2.8（上一版本）均引入了对共形预测区间的支持。这些更改包括新增的“conformal”置信方法，以及针对 Tibble（数据框）格式的预测结果展示优化，旨在帮助用户更好地理解在标准和嵌套式 Modeltime 预测工作流中，当前使用的置信方法及置信区间。\n\n#### 共形预测：\n\n- 将共形预测集成到嵌套式预测工作流中：`modeltime_nested_fit()` 和 `modeltime_nested_refit()`。#173\n- 更新了共形预测置信方法和置信区间的打印显示：\n  - `modeltime_forecast()`\n  - `extract_nested_test_forecast()`\n  - `extract_nested_future_forecast()`\n  - `modeltime_nested_forecast()`\n\n#### 其他更改：\n\n- Dials 参数：移除 `new_qual_param()` 内已弃用的 `default`。\n- 修复 dev-xregs 中的警告：在 `prepare_xreg_recipe_from_predictors()` 中使用 `all_of()`。\n- 修复损坏的测试：`test-tune_workflows` 中未使用的参数：`cores = 2`。","2023-12-08T15:50:40",{"id":171,"version":172,"summary_zh":173,"released_at":174},206542,"v1.2.8","# modeltime 1.2.8\n\n- 集成一致性预测。#173\n- 新示例：在 modeltime 中使用的一致性预测置信区间\n\n\n#### 其他变更：\n\n- 缩短了 CRAN 上的测试时间\n- CRAN 示例与测试：强制禁用并行核心，设置 `Sys.setenv(\"OMP_THREAD_LIMIT\" = 1)`\n- 将默认的并行处理核心数从所有可用核心（-1）更改为一个（1）：\n    - `control_refit()`\n    - `control_fit_workflowset()`\n    - `control_nested_fit()`\n    - `control_nested_refit()`\n    - `control_nested_forecast()`","2023-09-02T16:01:45",{"id":176,"version":177,"summary_zh":178,"released_at":179},206543,"v1.2.5","# modeltime 1.2.5\n\n- 修复平滑 `es()` 模型 #221","2023-02-08T14:03:30",{"id":181,"version":182,"summary_zh":183,"released_at":184},206544,"v1.2.4","修复 test-developer-tools-xregs.R 中失败的测试","2022-11-17T00:44:28",{"id":186,"version":187,"summary_zh":188,"released_at":189},206545,"v1.2.3","- 递归 `chunk_size`（性能优化）#197 #190\n- 递归模型修复 #194、#188、#187、#174\n- 新函数 `drop_modeltime_model` #160\n- 针对 `workflows` 模式 = \"regression\" 的更新","2022-10-18T17:37:05",{"id":191,"version":192,"summary_zh":193,"released_at":194},206546,"v1.2.2","# modeltime 1.2.2\n\n### 修复\n\n- 更新至 `hardhat 1.0.0` #182","2022-06-21T14:44:54",{"id":196,"version":197,"summary_zh":198,"released_at":199},206547,"v1.2.1","# modeltime 1.2.1\n\n### Trelliscope 绘图\n\n- `plot_modeltime_forecast()`: 公开 `facet_trelliscope()` 绘图参数。\n\n### 修复\n\n- 使用 `step_rm()` 移除日期列，而不是更新其角色 #181","2022-06-01T12:39:07",{"id":201,"version":202,"summary_zh":203,"released_at":204},206548,"v1.2.0","__新特性__\r\n\r\n许多绘图函数已升级，可与 `trelliscopejs` 配合使用，以便更轻松地可视化大量时间序列。\r\n\r\n- `plot_modeltime_forecast()`:\r\n    - 新增参数 `trelliscope`：用于可视化大量时间序列。\r\n    - 新增参数 `.facet_strip_remove`，用于移除分面条带，因为 `trelliscope` 会自动添加标签。\r\n    - 新增参数 `.facet_nrow`，用于配合 `trelliscope` 调整网格布局。\r\n    - 将默认参数 `facet_collapse = TRUE` 改为 `FALSE`，以更好地兼容 `Trelliscope JS`。这可能会导致某些图表中多个组占用条带的额外空间。","2022-04-07T19:48:35",{"id":206,"version":207,"summary_zh":208,"released_at":209},206549,"v1.1.1","## Fixes\r\n\r\n- Fixes issue of incorrect order of forecasts #142","2022-01-19T11:44:50",{"id":211,"version":212,"summary_zh":213,"released_at":214},206550,"v1.1.0","# Spark Backend\r\n\r\n- Modeltime now has a Spark Backend \r\n\r\n- [NEW Vignette - Modeltime Spark Backend](https:\u002F\u002Fbusiness-science.github.io\u002Fmodeltime\u002Farticles\u002Fmodeltime-spark.html) describing how to set up Modeltime with the Spark Backend. \r\n\r\n# New Algorithms: Smooth Package Integration\r\n\r\nIf users install `smooth`, the following models become available:\r\n\r\n- `adam_reg()`: Interfaces with the ADAM forecasting algorithm in `smooth`. \r\n\r\n- `exp_smoothing()`: A new engine \"smooth_es\" connects to the Exponential Smoothing algorithm in `smooth::es()`. This algorithm has several advantages, most importantly that it can use x-regs (unlike \"ets\" engine).\r\n\r\n# Nested Modeltime Improvements\r\n\r\n- New extractor: `extract_nested_modeltime_table()` - Extracts a nested modeltime table by row id. \r\n\r\n# Breaking Changes (potentially)\r\n\r\n- `extract_nested_train_split` and `extract_nested_test_split`: Changed parameter from `.data` to `.object` for consistency with other \"extract\" functions\r\n\r\n- Added a new logged feature to `modeltime_nested_fit()` to track the attribute \"metric_set\", which is needed for ensembles. Old nested modeltime objects will need to be re-run to get this new attribute. This will be used in ensembles. ","2021-10-18T16:29:38",{"id":216,"version":217,"summary_zh":218,"released_at":219},206551,"v1.0.0","# modeltime 1.0.0\r\n\r\n### New Feature: Nested (Iterative) Forecasting\r\n\r\n__Nested (Iterative) Forecasting__ is aimed at making it easier to perform forecasting that is traditionally done in a _for-loop_ with models like ARIMA, Prophet, and Exponential Smoothing. Functionality has been added to:\r\n\r\n#### Format data in a Nested Time Series structure\r\n\r\n- __Data Preparation Utilities:__ `extend_timeseries()`, `nest_timeseries()`, and `split_nested_timeseris()`.\r\n\r\n#### Nested Model Fitting (Train\u002FTest)\r\n\r\n- __`modeltime_nested_fit()`:__ Fits many models to nested time series data and organizes in a \"Nested Modeltime Table\". Logs Accuracy, Errors, and Test Forecasts. \r\n\r\n- __`control_nested_fit()`:__ Used to control the fitting process including verbosity and parallel processing. \r\n\r\n- __Logging Extractors:__ Functions that retrieve logged information from the initial fitting process. `extract_nested_test_accuracy()`, `extract_nested_error_report()`, and `extract_nested_test_forecast()`.\r\n\r\n#### Nested Model Selection\r\n\r\n- __`modeltime_nested_select_best()`__: Selects the best model for each time series ID. \r\n\r\n- __Logging Extractors:__ Functions that retrieve logged information from the model selection process. `extract_nested_best_model_report()`\r\n\r\n\r\n#### Nested Model Refitting (Actual Data)\r\n\r\n- __`modeltime_nested_refit()`:__ Refits to the `.future_data`. Logs Future Forecasts. \r\n\r\n- __`control_nested_refit()`:__ Used to control the re-fitting process including verbosity and parallel processing. \r\n\r\n- __Logging Extractors:__ Functions that retrieve logged information from the re-fitting process. `extract_nested_future_forecast()`.\r\n\r\n### New Vignette\r\n\r\n- [Nested Forecasting](https:\u002F\u002Fbusiness-science.github.io\u002Fmodeltime\u002Farticles\u002Fnested-forecasting.html) \r\n\r\n### Vignette Improvements\r\n\r\n- [Forecasting with Global Models](https:\u002F\u002Fbusiness-science.github.io\u002Fmodeltime\u002Farticles\u002Fmodeling-panel-data.html): Added more complete steps in the forecasting process so now user can see how to forecast each step from start to finish including future forecasting.  \r\n\r\n\r\n### New Accuracy Metric Set and Yardstick Functions\r\n\r\n- `extended_forecast_accuracy_metric_set()`: Adds the new MAAPE metric for handling intermittent data when MAPE returns Inf. \r\n- `maape()`: New yardstick metric that calculates \"Mean Arctangent Absolute Percentage Error\" (MAAPE). Used when MAPE returns Inf typically due to intermittent data. \r\n\r\n### Improvements\r\n\r\n- `modeltime_fit_workflowset()`: Improved handling of Workflowset Descriptions, which now match the `wflow_id`. ","2021-09-15T14:56:17",{"id":221,"version":222,"summary_zh":223,"released_at":224},206552,"v0.7.0","### Group-Wise Accuracy and Confidence Interval by Time Series ID\r\n\r\nWe've expanded Panel Data functionality to produce model accuracy and confidence interval estimates by a Time Series ID (#114). This is useful when you have a Global Model that produces forecasts for more than one time series. You can more easily obtain grouped accuracy and confidence interval estimates. \r\n\r\n* `modeltime_calibrate()`: Gains an `id` argument that is a quoted column name. This identifies that the residuals should be tracked by an time series identifier feature that indicates the time series groups. \r\n\r\n* `modeltime_accuracy()`: Gains a `acc_by_id` argument that is `TRUE`\u002F`FALSE`. If the data has been calibrated with `id`, then the user can return local model accuracy by the identifier column. The accuracy data frame will return a row for each combination of Model ID and Time Series ID. \r\n\r\n* `modeltime_forecast()`: Gains a `conf_by_id` argument that is `TRUE`\u002F`FALSE`. If the data has been calibrated with `id`, then the user can return local model confidence by the identifier column. The forecast data frame will return an extra column indicating the identifier column. The confidence intervals will be adjusted based on the local time series ID variance instead of the global model variance. \r\n\r\n### New Vignette\r\n\r\n[Forecasting Panel Data](https:\u002F\u002Fbusiness-science.github.io\u002Fmodeltime\u002Farticles\u002Fmodeling-panel-data.html) \r\n\r\n### New Algorithms\r\n\r\n#### THIEF: Temporal Hierarchical Forecasting\r\n\r\n- `temporal_hierarchy()`: Implements the `thief` package by Rob Hyndman and\r\nNikolaos Kourentzes for \"Temporal HIErarchical Forecasting\". #117\r\n\r\n### Bug Fixes\r\n\r\n- Issue #111: Fix bug with `modeltime_fit_workflowset()` where the workflowset (wflw_id) order was not maintained. \r\n","2021-07-16T11:34:04",{"id":226,"version":227,"summary_zh":228,"released_at":229},206553,"v0.6.1","__Parallel Processing__\r\n\r\n- New Vignette: [Parallel Processing](https:\u002F\u002Fbusiness-science.github.io\u002Fmodeltime\u002Farticles\u002Fparallel-processing.html)\r\n\r\n- `parallel_start()` and `parallel_stop()`: Helpers for setting up multicore processing. \r\n\r\n- `create_model_grid()`: Helper to generate model specifications with filled-in parameters from a parameter grid (e.g. `dials::grid_regular()`).\r\n\r\n- `control_refit()` and `control_fit_workflowset()`: Better printing. \r\n\r\n__Bug Fixes__\r\n\r\n- Issue #110: Fix bug with `cores > cores_available`.","2021-06-13T12:31:44",{"id":231,"version":232,"summary_zh":233,"released_at":234},206554,"v0.6.0","### Workflowset Integration\r\n\r\n`modeltime_fit_workflowset()` (#85) makes it easy to convert `workflow_set` objects to Modeltime Tables (`mdl_time_tbl`). Requires a refitting process that can now be performed in parallel or in sequence. \r\n\r\n\r\n### New Algorithms\r\n\r\n- CROSTON (#5, #98) - This is a new engine that has been added to `exp_smoothing()`. \r\n- THETA (#5, #93) - This is a new engine that has been added to `exp_smoothing()`.\r\n\r\n### New Dials Parameters\r\n\r\n`exp_smoothing()` gained 3 new tunable parameters:\r\n\r\n- `smooth_level()`: This is often called the \"alpha\" parameter used as the base level smoothing factor for exponential smoothing models.\r\n- `smooth_trend()`: This is often called the \"beta\" parameter used as the trend smoothing factor for exponential smoothing models.\r\n- `smooth_seasonal()`: This is often called the \"gamma\" parameter used as the seasonal smoothing factor for exponential smoothing models.\r\n\r\n### Parallel Processing\r\n\r\n- `modeltime_refit()`: supports parallel processing. See `control_refit()` \r\n- `modeltime_fit_workflowset()`: supports parallel processing. See `control_workflowset()` \r\n\r\n### Updates for parsnip >= 0.1.6\r\n\r\n- `boost_tree(mtry)`: Mapping switched from `colsample_bytree` to `colsample_bynode`. `prophet_boost()` and `arima_boost()` have been updated to reflect this change.  https:\u002F\u002Fgithub.com\u002Ftidymodels\u002Fparsnip\u002Fpull\u002F499\r\n\r\n### General Improvements\r\n\r\n- Improve Model Description of Recursive Models (#96)\r\n\r\n### Potential Breaking Changes\r\n\r\n- We've added new parameters to Exponential Smoothing Models. `exp_smoothing()` models produced in prior versions may require refitting with `modeltime_refit()` to upgrade their internals with the new parameters. \r\n","2021-05-30T11:45:32",{"id":236,"version":237,"summary_zh":238,"released_at":239},206555,"v0.5.1","# Modeltime 0.5.1 \r\n\r\n### Recursive Ensemble Predictions\r\n\r\n- Add support for `recursive()` for ensembles. ","2021-04-03T15:03:53",{"id":241,"version":242,"summary_zh":243,"released_at":244},206556,"v0.5.0","This release includes significant advances in forecasting with recursive panel data. \r\n\r\n### Recursive Predictions\r\n\r\n- `recursive()` (#71) - Received a full upgrade to work with Panel Data. \r\n- __New Vignette__: \"Recursive Forecasting\" with Modeltime\r\n\r\n### Breaking Changes\r\n\r\n- Deprecating `modeltime::metric_tweak()` for `yardstick::metric_tweak()`. The `yardstick::metric_tweak()` has a required `.name` argument in addition to `.fn`, which is needed for tuning. ","2021-03-31T09:40:33"]