【摘要】：隨著互聯(lián)網(wǎng)的快速發(fā)展和廣泛普及,伴隨著市場經(jīng)濟的逐漸成長和電子商務、物流行業(yè)等的迅猛增長,互聯(lián)網(wǎng)廣告得到了迅速的發(fā)展。機器學習的應用使得互聯(lián)網(wǎng)廣告發(fā)生了革命性的變化,在使用機器學習進行廣告CTR預估等工作中,特征調(diào)研是大家的主要工作,需要進行大量的實驗,這些實驗通�；诤Ａ康臄�(shù)據(jù)(數(shù)十或數(shù)百TB),大量而繁雜的大數(shù)據(jù)處理和結(jié)果分析工作存在著冗余,分散了調(diào)研人員的精力,使得調(diào)研效率難以提升。據(jù)此,基于Yii框架的模型特征分析管理平臺旨在減少冗余工作、提供快捷的實驗方式以及對實驗結(jié)果和分析數(shù)據(jù)的統(tǒng)一管理,提升調(diào)研效率。本文主要是從軟件工程的角度,描述分析、設(shè)計和開發(fā)模型特征分析管理平臺的完整過程,論文的主要工作體現(xiàn)在以下方面：(1)介紹了項目的啟動背景。平臺應對和解決的問題,主要是實驗任務配置和操作管理的界面化、數(shù)據(jù)的管理和可視化查看、對比等。(2)介紹了項目相關(guān)的技術(shù)。本項目基于Yii框架開發(fā),Yii使用MVC模式。論文中介紹了MVC設(shè)計模式和Yii框架,并描述了Yii框架的工作流程。機器學習是整個項目的應用背景,本文簡單的介紹了機器學習、模型訓練和特征評估等工作。FusionCharts是項目中用到的可視化開發(fā)包。(3) 詳細分析了項目的需求。詳細分析了用戶的工作流和數(shù)據(jù)流,給出了系統(tǒng)的主要用例。(4)描述了項目的總體設(shè)計和詳細設(shè)計。詳細描述了數(shù)據(jù)庫中重點數(shù)據(jù)表的設(shè)計,重點功能模塊的設(shè)計,比如用戶權(quán)限的管理、數(shù)據(jù)管理、全流程實驗的創(chuàng)建、特征分析、數(shù)據(jù)對比、平臺監(jiān)控管理等。(5)對項目的關(guān)鍵部分的實現(xiàn)做了詳細說明,給出了部分實現(xiàn)細節(jié)的代碼,以及實現(xiàn)結(jié)果的截圖等。
[Abstract]:With the rapid development and popularization of the Internet, with the gradual growth of the market economy and the rapid growth of e-commerce, logistics industry, Internet advertising has been rapidly developed. The application of machine learning has revolutionized the Internet advertising. In the work of using machine learning to predict advertising CTR, feature investigation is the main work of everyone, and a large number of experiments are needed. These experiments are usually based on a large amount of data (dozens or hundreds of TB), large and complicated big data processing and analysis of the results of the work there is redundancy, the distraction of the researchers, so that the efficiency of research is difficult to improve. Therefore, the model feature analysis management platform based on Yii framework aims to reduce redundant work, provide rapid experimental methods and unified management of experimental results and analysis data, and improve the efficiency of research. From the point of view of software engineering, this paper describes the complete process of analyzing, designing and developing the model feature analysis management platform. The main work of this paper is as follows: (1) the background of the project is introduced. The main problems to be solved by the platform are the interface of experiment task configuration and operation management, data management and visual view, contrast etc. (2) the project related technology is introduced. This project is based on the Yii framework to develop the Yii using MVC mode. This paper introduces MVC design pattern and Yii framework, and describes the workflow of Yii framework. Machine learning is the application background of the whole project. This paper briefly introduces the machine learning, model training and feature evaluation, and so on, which is the visual development kit used in the project. (3) the requirements of the project are analyzed in detail. The workflow and data flow of users are analyzed in detail, and the main use cases of the system are given. (4) the overall design and detailed design of the project are described. The design of key data tables in database, the design of key function modules, such as the management of user rights, the management of data, the creation of whole flow experiments, the analysis of features, the comparison of data, are described in detail. Platform monitoring and management. (5) the implementation of the key parts of the project is described in detail, and some of the implementation details of the code, as well as the implementation of the results of the screenshot.
【學位授予單位】：南京大學
【學位級別】：碩士
【學位授予年份】：2013
【分類號】：TP311.52

【參考文獻】

相關(guān)期刊論文 前1條

1 劉慶和;梁正友;;一種基于信息增益的特征優(yōu)化選擇方法[J];計算機工程與應用;2011年12期

，

本文編號：2232058