近年来为应对气候变化我国提出“双碳目标”，且国际油价动荡，柴油市场必受重大影响。加油站作为柴油市场的主要销售主体，提升其市场份额、创效能力是其零售业务利润的主要来源和保证，利用现代技术更科学地决策经营实现柴油的精准营销可有效提升其创效。大数据发展至今已经被广泛用于各行各业的精准营销场景，对于石油行业，加油站在多年的经营管理中积累了海量的数据，这些数据尚未被有效利用。且不同营销策略应用于不同的加油站其效果具有差异性。本研究旨在针对柴油营销活动，应用机器学习方法，对柴油车客户进站加油的影响因素进行量化分析，并根据营销反馈效果数据，科学的选择最适合应用该营销活动的加油站点进行营销，形成营销策略与加油站“一对一”的精准选择模式，从而使得收益最大化，为销售板块提供科学、智能的决策支撑。依托于A石油公司的管理销售大数据平台结合地理信息等数据对已经使用的特定营销方法的不同加油站效果进行量化分析，建立机器学习分类模型，对加油站是否适合进行柴油价格营销策略进行分类。主要结论：1.价格营销策略效果显著，可拉动柴油销量增幅达4倍以上；2.价格营销策略具有1-2月长尾效应；3. 不同地区营销效果存在差异；4.不同道路类型加油站营销效果差异显著；5. 随机森林为构建柴油价格直降营销策略站点选择模型最优算法，测试集准确率为73.4%，效果优于随机森林模型和KNN算法模型；6.对于加油站柴油价格直降精准营销，加油站活动前日均柴油销量、所在省份、到道路的最短距离对其影响排名前三，加油站附近道路月平均车流量、所在商圈类型、周边竟对个数也有影响。

In recent years, in order to deal with climate change, China has put forward the "double carbon target", and the international oil price is volatile, the diesel market will be greatly affected. As the front-line terminal of the diesel market, gas stations improve their efficiency and market share, which is the main source and guarantee of their retail business profits. Using contemporary technology to make more scientific decision-making and operation to realize the precise marketing of diesel can effectively improve their efficiency. Since its development, big data has been widely used in precision marketing scenarios in all walks of life. For the oil industry, gas stations have accumulated a large amount of operation data in many years of operation, which have not yet been utilized. And different marketing strategies are applied to different gas stations, and their effects are different. The purpose of this study is to quantitatively analyze the influencing factors of diesel vehicle customers' refuelling in the station by using machine learning method for diesel marketing activities, and scientifically select the gas station most suitable for the marketing activities according to the marketing feedback effect data, so as to form a precise selection mode of marketing strategy and gas station "one-to-one", so as to maximize the income and provide scientific and intelligent decision support for the sales sector. Relying on the management and sales big data platform of a petroleum company, combined with geographic information and other data, this paper makes a quantitative analysis on the effects of different gas stations with specific marketing methods, establishes a machine learning classification model, and classifies whether the gas station is suitable for diesel price marketing strategy. Main conclusions: 1. The effect of price marketing strategy is remarkable, which can drive the increase of diesel sales by more than 4 times; 2. The price marketing strategy has a long tail effect from January to February; 3. There are differences in marketing effects in different regions; 4. Different types of gas stations in the county have different marketing efficiency; 5. Random forest is the best algorithm for constructing the site selection model of diesel price straight down marketing strategy. The accuracy of the test set is 73.4%, which is better than random forest model and KNN algorithm model; 6. For the precise marketing of the straight drop of diesel prices in gas stations, the average daily diesel sales before the gas station activity, the province where it is located and the shortest distance to the road rank the top three. The average monthly traffic flow of the roads near the gas station, the type of business district where it is located and the number of surrounding areas also have an impact.