2015-2020 美国市场牛油果价格分析和预测

作为一位牛油果资深爱好者，苦于牛油果价格实在太贵，Sam 店一个品相俱佳的大果要8块钱，鲜果一号之类等其他精品水果店居然有的要卖到10+一个，还是中果，硬邦邦，不怎么好吃的那种，简直黑心。

同时呢，作为一位申请美帝研究生的大四狗，我严肃认真地要把牛油果价格因素纳入到我选择的范围内。同时，我真的很好奇，美帝牛油果的价格是咋样的。当然最重要的是要完成当前303大数据作业🧐，因为report有点写不出来，就先发个专栏整理整理思路。

以下分析以及可视化，借鉴了之前大神们的分析：（他们做的图真的好好看啊！）

https://www.kaggle.com/janiobachmann/price-of-avocados-pattern-recognition-analysis

但是他们的数据集是从2015-2018，已经有一些年份了，不能如实反应当今市场，尤其是新冠后的影响。碰巧有大神更新了数据集，在原有的基础上新增到2020.5月的牛油果数据。利用前人经验，做未来预测分析

https://www.kaggle.com/timmate/avocado-prices-2020

注：本篇所有分析预测都在r上完成，95%借助ggplot2，yysy，r做图真的好看。预测部分，偷了点懒，直接用的是Facebook开发的prophet包，因为之前没有接触过时间相关的预测，LSTM目前还没有学到家，但ddl在前，等有空再用python补上。

下载并读取

df <- read.csv("/Users/yuxuan/Desktop/INT301-Avocado-prediction/avocado-updated-2020.csv")
head(df)

其中以下是我们这次的实验对象 - date - 观察的时间 - average_price - 每一个牛油果的平均价格 - total_volume - 当日售出了多少牛油果 - year - 年份（Date格式） - type - 种类，是有机的还是普通的 - geography - 数据来源的地区

至于X4046,X4225,X4770代表的是牛油果的种类(PLU code)都是Hass 牛油果，只是大小有区别

检查是否存在缺失（missing value）

sum(is.na(df))

发现数据集无缺失值

根据种类做density plot

library(ggplot2)
options(repr.plot.width = 8, repr.plot.height = 4)
density_plot <- ggplot(df, aes(x=average_price, fill=type))+
  geom_density()+
  facet_wrap(~type)+
  theme_minimal()+
  theme(plot.title = element_text(hjust = 0.5),legend.position = "bottom")+
  labs(title = "Avocado Price by type")+
  scale_fill_brewer(palette = "Set2")
density_plot

图中可以发现

• 普通的牛油果（绿色）大部分集中在1$ 附近，形状也比较高瘦，值域区间相比之下也比较小
• 但有机的牛油果（红色）则显得更加敦实，值域也宽，最贵的居然要卖到3$

量化具体的一些比率

library(dplyr)
vol_type <- df %>% group_by(type) %>% summarise(average_volume = round(mean(total_volume),3),average_price = round(mean(average_price),3)) %>% mutate(volume_percent= round(prop.table(average_volume)*100,3))
vol_type

Type	Average Volume	Average Price	Volume percent
Conventional	1,818,206 (1.8 M)	1.16 $	96.8%
Organic	60,127 (0.06 M)	1.62 $	3.2%

从表格中我们可以发现

• 销售的普通牛油果在市场以均价1.16$（7.58¥）占比居然高达97%
• 相比之下有机牛油果1.62$ (10.58¥) 市场占比大概只有3%

2015-2020间普通和有机牛油果的价格波动曲线

df$date <- as.Date(df$date, "%Y-%m-%d")

## Sort the dates and order the datesets in date
df <- df[order(df$date),]

## Make the plot
comparision_plot <- df %>% select(date, average_price, type) %>%
  ggplot(aes(x=date,y=average_price))+
  geom_area(aes(color=type,fill=type),alpha=0.3,position=position_dodge(0.8))+
  theme_bw()+
  scale_color_manual(values = c("#ED7921","#62BE51"))+
  scale_fill_manual(values = c("#FD833E","#B8FC5F")
)

comparision_plot

ggplot(data=df, aes(x=date, y=average_price,col=type))+
  geom_line()+
  facet_wrap(~ type)+
  theme_bw()+
  theme(legend.position = "position")

• 有机的价格永远要高于普通的
• 价格呈现某种季节性的波动，符合水果季节性波动的常识
• 是否和销售的量呈现关联，下文探索

2015-2020年间的牛油果总的价格分布

ggplot(seasonal_df,aes(x=average_price,fill=as.factor(year)))+
geom_density(alpha=0.5)+
theme_economist()+
facet_wrap(~year)+
theme(plot.title = element_text(hjust = 0.5),plot.background = element_rect(fill="#D5D8DC"))+
guides(fill=FALSE)+
labs(title = "Distribution of Prices by year",x='Average Price',y='Density')+
scale_fill_manual(values = c("#DA4511", "#FFBD00", "#6A953F", "#9A6233", "#D3AE7C", "#307CA1"))

• 6年间的价格分布，其中2017年最成正态分布的形状，高端和低端牛油果都在市场分一杯羹

价格波动曲线和牛油果售卖的量关系

library(ggplot2)
library(ggthemes)
library(cowplot)

options(repr.plot.width=8, repr.plot.height=6)

## average-price with time series
conventional_monthly <- conventional %>%
    ggplot(aes(x=date,y=average_price))+
    geom_line(color="#5C374C")+
    theme_economist()+
    theme(plot.title = element_text(hjust = 0.5),plot.background = element_rect(fill = "#D5D8DC"))+
    labs(title = "Conventional Avocados")+
    geom_hline(yintercept = max(conventional$average_price),linetype="dashed",color = "red")+
    geom_hline(yintercept = min(conventional$average_price),linetype="dashed",color = "blue")

organic_monthly <- organic %>%
    ggplot(aes(x=date,y=average_price))+
    geom_line(color="#58D68D")+
    theme_economist()+
    theme(plot.title = element_text(hjust = 0.5),plot.background = element_rect(fill = "#D5D8DC"))+
    labs(title = "Organic Avocados")+
    geom_hline(yintercept = max(organic$average_price),linetype="dashed",color = "red")+
    geom_hline(yintercept = min(organic$average_price),linetype="dashed",color = "blue")

## create a volume chart
conventional_volume <- conventional %>%
    ggplot(aes(x=date,y=total_volume))+
    geom_bar(stat = 'identity',fill="#7FB3D5",color="black")+
    theme_economist()+
    theme(plot.title = element_text(hjust = 0.5),plot.background = element_rect(fill = "#D5D8DC"))+
    geom_smooth(method = "loess",color="red")

organic_volume <- organic %>%
    ggplot(aes(x=date,y=total_volume))+
    geom_bar(stat = 'identity',fill='#58D68D',color="black")+
    theme_economist()+
    theme(plot.title = element_text(hjust = 0.5),plot.background = element_rect(fill = "#D5D8DC"))+
    geom_smooth(method = "loess",color ="red")

plot_grid(conventional_monthly,organic_monthly,conventional_volume,organic_volume,nrow = 2,ncol = 2,labels = c("A","B","C","D"))

• 为了找到季节性规律（seasonal patterns）我将平均售价和销售量以月为单位进行统计分析
• A,B图代表的都是以月为单位的平均销售价格（monthly），蓝线代表最小值（min）红线代表最大值（max）
• 普通牛油果最贵的一个月也就1.8$，最便宜的一个月0.82$；有机的最贵要2.1$，最便宜的也要1.21$
• C,D代表的是以月为单位的销量，红线代表的是趋势
• 美帝人民对牛油果的爱是一贯的，销量呈逐年上升的趋势，这里指的是普通牛油果
• 19-20年，可能由于经济形势下滑以及20年后的新冠疫情，对有机牛油果的需求减少，未能像普通牛油果一样持续增长
• 月销量也呈现某种季节性规律，需要接下来的仔细分析

2015-2020年间以月为单位比较销量和价格

## Process the data into year and month format
library(forecast)
seasonal_df <- read.csv("/Users/yuxuan/Desktop/INT303-Avocado-prediction/avocado-updated-2020.csv")
seasonal_df$month_year <- format(as.Date(seasonal_df$date),"%Y-%m")
seasonal_df$month <- format(as.Date(seasonal_df$date),"%m")

## Change the month from a Date format into a numerical foramt, then convert to the three letter format
seasonal_df$monthabb <- sapply(seasonal_df$month, function (x) month.abb[as.numeric(x)])
seasonal_df$monthabb <- factor(seasonal_df$monthabb,levels=month.abb)
seasonal_df$monthabb <- factor(seasonal_df$monthabb)

## Set the figure size
options(repr.plot.width=10,repr.plot.height=8)

## Analyze the price by month

conv_price <- seasonal_df %>% select(type,year,monthabb,average_price) %>% filter(type=="conventional") %>% group_by(year,monthabb) %>% summarise(avg=mean(average_price))

org_price <- seasonal_df %>% select(type,year,monthabb,average_price) %>% filter(type=="organic") %>% group_by(year,monthabb) %>% summarise(avg=mean(average_price))

conv_price <- ts(conv_price$avg,start = 2015,frequency = 12)
org_price <- ts(org_price$avg,start = 2015,frequency = 12)

## Analyze the volume by month
conv_volume <- seasonal_df %>% select(type,year,monthabb,total_volume) %>% filter(type=="conventional") %>% group_by(year,monthabb) %>% summarise(avg=mean(total_volume))

org_volume <- seasonal_df %>% select(type,year,monthabb,total_volume) %>% filter(type=="organic") %>% group_by(year,monthabb) %>% summarise(avg=mean(total_volume))

conv_volume <- ts(conv_volume$avg,start = 2015,frequency = 12)
org_volume <- ts(org_volume$avg,start = 2015,frequency = 12)

byyearplot_price_conv <- ggseasonplot(conv_price,year.labels = TRUE,year.labels.left = TRUE)+
theme_economist()+
theme(plot.title = element_text(hjust = 0.5),plot.background = element_rect(fill="#D5D8DC"))+
labs(title = "Average conventional Avocados price \n by year for each month", y="Average Price")+
scale_fill_manual(values = c("#922B21", "#EE865D", "#DDCD5E", "#59BEC4", "#048B9F", "#114676"))

byyearplot_price_org <- ggseasonplot(org_price,year.labels = TRUE,year.labels.left = TRUE)+
theme_economist()+
theme(plot.title = element_text(hjust = 0.5),plot.background = element_rect(fill="#D5D8DC"))+
labs(title = "Average organic Avocados price \n by year for each month", y="Average Price")+
scale_fill_manual(values = c("#922B21", "#EE865D", "#DDCD5E", "#59BEC4", "#048B9F", "#114676"))

byyearplot_volume_conv <- ggseasonplot(conv_volume,year.labels = TRUE,year.labels.left = TRUE)+
theme_economist()+
theme(plot.title = element_text(hjust = 0.5),plot.background = element_rect(fill="#D5D8DC"))+
labs(title = "Average conventional Avocados volume \n by year for each month", y="Average volume")+
scale_fill_manual(values = c("#922B21", "#EE865D", "#DDCD5E", "#59BEC4", "#048B9F", "#114676"))

byyearplot_volume_org <- ggseasonplot(org_volume,year.labels = TRUE,year.labels.left = TRUE)+
theme_economist()+
theme(plot.title = element_text(hjust = 0.5),plot.background = element_rect(fill="#D5D8DC"))+
labs(title = "Average organic Avocados volume by year \n for each month", y="Average volume")+
scale_fill_manual(values = c("#922B21", "#EE865D", "#DDCD5E", "#59BEC4", "#048B9F", "#114676"))

plot_grid(byyearplot_price_conv,byyearplot_price_org,byyearplot_volume_conv,byyearplot_volume_org,nrow = 2,ncol = 2,labels = c("A","B","C","D"))

options(repr.plot.width=10,repr.plot.height=8)
conv_patterns <- seasonal_df %>% select(monthabb,average_price,type) %>% filter(type=="conventional") %>% group_by(monthabb) %>% summarise(avg=mean(average_price)) %>%
  ggplot(aes(x=monthabb, y=avg))+
  geom_point(color="#F35D5D",aes(size=avg))+
  geom_line(group=0)+
  theme_economist()+
  theme(legend.position = "none",plot.title = element_text(hjust = 0.5),plot.background = element_rect(fill="#D5D8DC"))+
  labs(title = "Conventional Avocados",x="Month",y="Average Price")

organic_patterns <- seasonal_df %>% select(monthabb,average_price,type) %>% filter(type=="organic") %>% group_by(monthabb) %>% summarise(avg=mean(average_price)) %>%
ggplot(aes(x=monthabb,y=avg))+
  geom_point(color="#F35D5D",aes(size=avg))+
  geom_line(group=0)+
  theme_economist()+
  theme(legend.position = "none",plot.title = element_text(hjust = 0.5),plot.background = element_rect(fill="#D5D8DC"))+
  labs(title = "Organic Avocados",x="Month",y="Average Price")

whole_patterns <- seasonal_df %>% select(monthabb,average_price,type) %>% group_by(monthabb) %>% summarise(avg=mean(average_price)) %>%
  ggplot(aes(x=monthabb,y=avg))+
  geom_point(color="#F35D5D",aes(size=avg))+
  geom_line(group=0)+
  theme_economist()+
  theme(legend.position = "none",plot.title = element_text(hjust = 0.5),plot.background = element_rect(fill="#D5D8DC"))+
  labs(title = "All Avocados",x="Month",y="Average Price")


plot_grid(conv_patterns,organic_patterns,whole_patterns,nrow = 3)

conv_patterns_vol <- seasonal_df %>% select(monthabb,total_volume,type) %>% filter(type=="conventional") %>% group_by(monthabb) %>% summarise(avg=mean(total_volume)) %>%
  ggplot(aes(x=monthabb, y=avg))+
  geom_point(color="#F35D5D",aes(size=avg))+
  geom_line(group=0)+
  theme_economist()+
  theme(legend.position = "none",plot.title = element_text(hjust = 0.5),plot.background = element_rect(fill="#D5D8DC"))+
  labs(title = "Conventional Avocados volume",x="Month",y="Average Price")

organic_patterns_vol <- seasonal_df %>% select(monthabb,total_volume,type) %>% filter(type=="organic") %>% group_by(monthabb) %>% summarise(avg=mean(total_volume)) %>%
ggplot(aes(x=monthabb,y=avg))+
  geom_point(color="#F35D5D",aes(size=avg))+
  geom_line(group=0)+
  theme_economist()+
  theme(legend.position = "none",plot.title = element_text(hjust = 0.5),plot.background = element_rect(fill="#D5D8DC"))+
  labs(title = "Organic Avocados volume",x="Month",y="Average Price")

whole_patterns_vol <- seasonal_df %>% select(monthabb,total_volume,type) %>% group_by(monthabb) %>% summarise(avg=mean(total_volume)) %>%
  ggplot(aes(x=monthabb,y=avg))+
  geom_point(color="#F35D5D",aes(size=avg))+
  geom_line(group=0)+
  theme_economist()+
  theme(legend.position = "none",plot.title = element_text(hjust = 0.5),plot.background = element_rect(fill="#D5D8DC"))+
  labs(title = "All Avocados volume",x="Month",y="Average Price")


plot_grid(conv_patterns_vol,organic_patterns_vol,whole_patterns_vol,nrow = 3)

• 综合来说，每年的9，10 月份牛油果的平均价格会达到一年中的最高值，2月达到一年的最低值
• 对于销量来说，美国人喜欢在2月和5月购买牛油果，11月份买牛油果的意愿最低
• 查资料可得，牛油果一般在8，9月份成熟收获，加上采摘，包装和运输时间所以会有一定的延迟，可以看见图中的八月都是呈现上升趋势。
• 可以发现销售量和价格呈现一定的负相关，这符合我们的常识，人们喜欢在价格低的时候购买，而一样水果或者蔬菜则会在刚上市时价格逐步上升，达到最高，然后再下降
• 为什么随着采摘供应和价格会成一定的正相关？我的猜测是，就像苏州人说的“时鲜货”，刚刚摘下的牛油果肯定是又大又好有新鲜，而且前一年的量已经消耗的差不多了，所以价格会有一阵子的上涨，然后下探，符合图中的趋势走向
• 也可以清晰的看见，销量的确是逐年增加的，大概率是网红不遗余力的宣传，将牛油果作为健康活力，fashion的代名词。这路子的确是正确的，因为美国人民消耗的牛油果总的来说逐年递增

2015-2020 以季节为单位分析牛油果价格和销量

options(repr.plot.width=10,repr.plot.height=8)

## seperate the month into four seasons
seasonal_df$season <- ifelse(seasonal_df$month %in% c("03","04","05"),"Spring",
ifelse(seasonal_df$month %in% c("06","07","08"),"Summer",
ifelse(seasonal_df$month %in% c("09","10","11"),"Autumn","Winter")))

## Prepare to analyze the results
seasonality_plot_conventional_price <- seasonal_df %>% select(season,year,average_price,type) %>% filter(type =="conventional") %>% group_by(season,year) %>% summarise(avg=mean(average_price)) %>%
  ggplot(aes(x=season,y=avg,color=season))+
  geom_segment(aes(x=season,xend=season,y=0,yend=avg),show.legend = FALSE)+
  coord_flip()+
  facet_wrap(~as.factor(year))+
  theme_minimal()+
  theme(plot.title = element_text(hjust = 0.5),plot.background = element_rect(fill="#F4F6F7"))+
  labs(title = "Conventional Avocados average price by Season",x="",y="Average price")+
  geom_text(aes(x=season,y=0.01,label=paste0("$ ",round(avg,2))),hjust=-0.5,vjust=-0.5,size=4,color="black",fontface='italic',angle=360)

seasonality_plot_conventional_volume <- seasonal_df %>% select(season,year,total_volume,type) %>% filter(type=="conventional") %>% group_by(season,year) %>% summarise(avg=round(mean(total_volume/1000000),2)) %>%
  ggplot(aes(x=season,y=avg,color=season))+
  geom_segment(aes(x=season,xend=season,y=0,yend=avg),show.legend = FALSE)+
  coord_flip()+
  facet_wrap(~as.factor(year))+
  theme_minimal()+
  theme(plot.title = element_text(hjust = 0.5),plot.background = element_rect(fill="#F4F6F7"))+
  labs(title = "Conventional Avocados total volume by Season",x="",y="Average volume")+
  geom_text(aes(x=season,y=0.01,label=paste0(avg," m")),hjust=-0.5,vjust=-0.5,size=4,color="black",fontface='italic',angle=360)

#plot_grid(seasonality_plot_conventional_price,seasonality_plot_conventional_volume,nrow = 2)

seasonality_plot_organic_price <- seasonal_df %>% select(season,year,average_price,type) %>% filter(type =="organic") %>% group_by(season,year) %>% summarise(avg=mean(average_price)) %>%
  ggplot(aes(x=season,y=avg,color=season))+
  geom_segment(aes(x=season,xend=season,y=0,yend=avg),show.legend = FALSE)+
  coord_flip()+
  facet_wrap(~as.factor(year))+
  theme_minimal()+
  theme(plot.title = element_text(hjust = 0.5),plot.background = element_rect(fill="#F4F6F7"))+
  labs(title = "Organic Avocados average price by Season",x="",y="Average price")+
  geom_text(aes(x=season,y=0.01,label=paste0("$ ",round(avg,2))),hjust=-0.5,vjust=-0.5,size=4,color="black",fontface='italic',angle=360)

seasonality_plot_organic_volume <- seasonal_df %>% select(season,year,total_volume,type) %>% filter(type=="organic") %>% group_by(season,year) %>% summarise(avg=round(mean(total_volume/1000000),2)) %>%
  ggplot(aes(x=season,y=avg,color=season))+
  geom_segment(aes(x=season,xend=season,y=0,yend=avg),show.legend = FALSE)+
  coord_flip()+
  facet_wrap(~as.factor(year))+
  theme_minimal()+
  theme(plot.title = element_text(hjust = 0.5),plot.background = element_rect(fill="#F4F6F7"))+
  labs(title = "Organic Avocados total volume by Season",x="",y="Average volume")+
  geom_text(aes(x=season,y=0.01,label=paste0(avg," m")),hjust=-0.5,vjust=-0.5,size=4,color="black",fontface='italic',angle=360)

plot_grid(seasonality_plot_conventional_price,seasonality_plot_organic_price,seasonality_plot_conventional_volume,seasonality_plot_organic_volume,nrow = 2,ncol = 2,labels = c("A","B","C","D"))

• 春（3-5），夏（6-8），秋（9-11），冬（12-2）
• A,B 代表的是牛油果的平均价格，以有机和非有机划分；C,D 代表的则是销量，同样以有机和非有机划分，m代表million百万
• 总的来说，春冬季买牛油果比较划算，均价最低，同样也反应在了销量上，春夏销量最高，因为牛油果自从8，9月份成熟以后已经大量充斥了市场，所以价格会比较低，吸引较多的买家，直到下一批牛油果的成熟

牛油果价格和城市的关系

library(forcats)
geo_conv_price <- seasonal_df %>% select(geography,average_price,type) %>% filter(type=="conventional") %>% group_by(geography)%>%  summarise(avg=round(mean(average_price),2)) %>% arrange(avg) %>% slice(1:6)  %>% mutate(geography=fct_reorder(geography,desc(avg))) %>%
  ggplot(aes(x=geography,y=avg))+
  geom_bar(stat = "identity",position = "dodge",alpha=.6,width =.4,show.legend = FALSE)+
  coord_flip()+
  geom_text(aes(label=avg),vjust=1.5,color='black',size=5)+
  theme_minimal()+
  theme(title = element_text(hjust = 0.6),plot.background = element_rect(fill="#F4F6F7"))+
  labs(title = "Conventional Avocados average price by geography",x="Geography place",y="Average Price")

geo_org_price <- seasonal_df %>% select(geography,average_price,type) %>% filter(type=="organic") %>% group_by(geography)%>%  summarise(avg=round(mean(average_price),2)) %>% arrange(avg) %>% slice(1:6) %>% mutate(geography=fct_reorder(geography,desc(avg))) %>%
  ggplot(aes(x=geography,y=avg))+
  geom_bar(stat = "identity",position = "dodge",alpha=.6,width =.4,show.legend = FALSE)+
  coord_flip()+
  geom_text(aes(label=avg),vjust=1.5,color='black',size=5)+
  theme_minimal()+
  theme(title = element_text(hjust = 0.6),plot.background = element_rect(fill="#F4F6F7"))+
  labs(title = "Organic Avocados average price by geography",x="Geography place",y="Average Price")

plot_grid(geo_conv_price,geo_org_price,nrow = 2)

• 通过计算返回前六低的地区和城市
• 以非有机为例：菲尼克斯（凤凰城），休斯顿，达拉斯，中南部城市群，新墨西哥洲，Nashville（靠近印第安纳洲）
• 通过地图查询，前五个都在美国的中西部，靠近墨西哥
• 通过查询，牛油果原产地墨西哥，美国价格最低的前5个城市很有可能是种植牛油果的基地，所以牛油果价格便宜
• 维基百科上说之前加州和佛罗里达是牛油果的主产地，但随着水资源的价格上涨，牛油果产地发生了写偏移，我合理怀疑是向着上述地点前一，前五个城市，大致都分布在同一个纬度上
• 下面的链接说的是美国现阶段牛油果种植地方，提到了圣安东尼奥，也是中西部的城市，毗邻休斯顿，达拉斯，因为气候很像墨西哥，适合牛油果种植，并且淡水相对便宜

https://www.hunker.com/12003820/in-what-areas-of-the-united-states-can-you-grow-an-avocado-treewww.hunker.com

• 这里我用Google map将上述六个城市给标注出来了，的确是近似处于同一纬度，牛油果前五便宜的城市的确靠近墨西哥

牛油果价格预测和走势图

Quick Start

library(prophet)
library(cowplot)
library(gridExtra)
library(ggplot2)
library(ggpubr)
library(devtools)
## Select and filter the data and fit the model
price_predict <- df %>% select(date,average_price) %>% group_by(date) %>% summarise(avg=mean(average_price))
colnames(price_predict) <- c('ds','y')
model <- prophet(price_predict,daily.seasonality = TRUE)

## make prediction or forecast the results
future <- make_future_dataframe(model,periods = 365)
forecast <- predict(model,future)
plot(model,forecast)+
theme_minimal()+
  theme(plot.title = element_text(hjust = 0.5),plot.background = element_rect(fill="#F4F6F7"))+
  labs(title = "Avocados price prediction in next year",x="year",y="Average Price")

trend <- prophet_plot_components(model,forecast)

year_trend <- trend[1]
month_trend <- trend[3]
week_trend <- trend[2]
time_trend <- trend[4]

overall_trend <- c(year_trend,month_trend,week_trend,time_trend)
ggarrange(plotlist = overall_trend, labels = c('A', 'B','C','D'))

• 利用Facebook开发的prophet包进行时间序列上的价格预测
• 可以看见价格还是会呈现一个季节性波动，但是价格可能会走低
• 联系到当前新冠疫情在美国的疯狂爆发，牛油果销量下滑是必然的，走势向下也符合预期

• prophet包自带的功能，可以根据时间，细化到每一天以小时为单位的时间预测
• A图是以年为单位的分析预测：从15年开始，牛油果的价格一直保持正增长，17年是疯狂的一年，以高于1.30的幅度快速增长，这也应和了我的个人感受，因为我就是17年才知道这种水果，也就是说网红公司在这一年大幅度的疯狂宣传这款水果，然后在价格上炒的越来愈高。18年后价格增速趋于稳定，稳定在每年1.28左右增长；至于未来的一年，prophet也给出了自己的预测，区间大概在[1.20-1.34]。考虑到北美新冠疫情的影响，我觉得很有可能会以1.20左右，但仍然是正增长，据我在推特上观察，北美千禧年一代对牛油果仍然是非常非常喜欢，各种可爱有趣的漫画是层出不穷。我相信2020年牛油果的价格仍然上涨，只是增速放缓
• B图是以月为单位的分析，可以看见在5月份，价格转为正增长，持续到10月份，价格达到峰值，随后价格开始下跌，到来年的2月份达到低点，符合之前的分析
• C图是以每周为单位分析，牛油果在周末价格最高，符合美国家庭周末在周末进行大采购的习惯，所以那两天，价格也最高
• D图是以每天来分析，没有参考意义，因为我的数据集就没有这个带小时为单位的时间

完整版的代码请在GitHub下载

https://github.com/yuxuanwu17/INT303-Avocado-prediction