SAS：预测未来销售额的简单方法

Question

我有这样的数据：

>   Data sales;
>     Infile datalines missover;
>     Input year $ month  sales;
>     Datalines;
>     2016 1 489
>     2016 2 475
>     2016 3 523
>     2016 4 546
>     2016 5 675
>     2016 6 667
>     2016 7 334
>     2016 8 359
>     2016 9 521
>     2016 10 526
>     2016 11 459
>     2016 12 439
>     2017 1 500
>     2017 2 445
>     2017 3 531
>     2017 4 541
>     2017 5 620
>     2017 6 630
>     2017 7 300
>     2017 8 350
>     2017 9 502
>     2017 10 524
>     2017 11 402
>     2017 12 469
>     ;

例如，我希望采用去年的平均增长，并用它来预测接下来 12 个月的月份。我不是在寻找高级方法，这纯粹是为我编程。

到目前为止，我有这个（太丑了）：

    data sales2;
set sales;
lag_sales=lag(sales);
diff=sales-lag_sales;
start_month=month(intnx("month",today(),-12));
start_year=year(intnx("year",today(),-1));
if year>=start_year and month>=start_month;
run;
proc sql;
create table sales3 as
select avg(diff) as avg_diff
from sales2 ; quit;
proc sql;
create table sales4 as
select year, month, monotonic() as row_no, sales, avg_diff,
       sales+avg_diff*monotonic() as sales_pred
from sales2, sales3
;
quit;

Answer 1

下面的代码回答了您的问题/您正在尝试做什么。请注意，我使用了：

保留方法而不是滞后函数，
改为观察号 montonic().

如果您有很多记录，这将提高您的查询性能

请注意，您还可以根据移动平均线预测销售额，而不仅仅是差值（未包含在代码中）。

基于您的销售数据集的代码：

/*Calculate the difference*/
data diff;
set sales;
row = _n_;
retain prev;
if row=1 then do; prev=0; diff=0; output; prev=sales;end;
else do; diff=sales -prev; output; prev=sales; end;
run;
/*select avg for 2016, for live data feed change to -1*/
proc sql noprint;
select avg(diff) into:avg from diff where input(year,4.)=year(today())-2;
quit;
/*select avg for 2017, for live data feed remove the -1*/
proc sql;
create table sales_new as
select year,month, sales, &avg as avg_diff, Sales+&avg as sales_pred from diff where input(year,4.)=year(today())-1;
quit;

差异Table:

输出：

SAS：预测未来销售额的简单方法

SAS: simple way to predict future sales

sas

predict