|
x**********0 发帖数: 163 | 2 我看过arima,可是不太明白
test stationary, 是test 每一个variable 的stationary吧,然后再看model
residual的stationary吗? |
|
g**u 发帖数: 205 | 3 可以unit root test啊。我记得proc arima就能做到呢。 |
|
m*******1 发帖数: 855 | 4 有阿,ARIMA, exponential, 我们学了好几个,我都忘记了~~ 哪位高人出来指点指点
,我也复习和学习一下 |
|
y**3 发帖数: 267 | 5 I have only 7 yearly data points for a time series. Is it possible to do a
time series analysis to predict 8 year 9yrtc?
For a Arima model, minimum 40-50 data points is required.
Is there any expert with time series can help?\
thanks |
|
o**n 发帖数: 22 | 6 Please contact me through ZHAN NEI XIN XIANG.Let me know your background (
time serires, regression analysis, and multivariate statistics background
preferred) and the estimated time & cost. Thanks.
I will require a 12-15 page report (double-spaced, inclusive of diagram)
based on the following topic. The report is split into three sections:
1) A nonlinear time series of unknown origin (approx t=4000) is provided.
Based on standard analysis plot, tell me all you can about the time series
of unkno... 阅读全帖 |
|
X*****O 发帖数: 78 | 7 在做一个SKU Unit的time series monthly forecasting project,用了Winters,
Exponential Smoothing, ARIMA等model之后,效果不怎么样,R-square在0.5左右,
MAPE也较高。明显model不fit...请版上大牛指教在建model过程中是哪里出了问题,多
谢多谢 |
|
|
a***d 发帖数: 336 | 9 if you have good exogenous variables, that will probably help model accuracy
a lot. |
|
s*********h 发帖数: 6288 | 10 在工作中需要了解到什么程度?
ARIMA
SARIMA
ARAR
Holt-Winters smoothing
ARCH
GARCH
还有什么基本要素需要了解的嘛? |
|
c******y 发帖数: 17 | 11 This position is responsible for providing call volume forecasting to
critical marketing lines of business including Balance Transfer, Reissue,
Usage and etc. The candidate will be responsible for evaluating existing
models as well as creating new ones in those areas. There will also be a
heavy emphasis on providing analytics support to the work force management
team to help drive key business decisions.
New products and services are continually introduced and this position is
also required to o... 阅读全帖 |
|
C*****X 发帖数: 5 | 12 大家好!
我们Cipher Ground (www.cipherground.org)正在筹建Cipher Finance项目公司,与国
内一知名Quant平台合作,从事量化分析服务,面向国内160多家股票、期货公司。
我们需要以下人才
理论人才:数学、统计、云计算、金融;
实践人才:具有对冲基金实战经验的人才
具体技术:统计模型:ARIMA、GARCH;神经网络;SVM;分形;小波分析/谱分析;隐马
尔科夫;人工智能
报酬:根据起步期(3个月)的贡献,分配相应Cipher Finance公司股份
工作地点:不限(由于是数据分析,不需要固定地点)
若感兴趣,请将问题或者CV发至:
i**[email protected]
谢谢!
Cipher Finance |
|
|
A*******s 发帖数: 3942 | 14 my 2 cents--really depends on purpose
for pure prediction, i prefer quick, dirty & violent data mining--
1. do feature engineering first, generate as many predictors as possible,
for example, lags, diffs, moving averages
2. through all of them into machine learning models
for simulation purpose, then we need to use classic time series model, ARIMA
, GARCH or whatever textbook model that has good theorectical support.
indexing |
|
v*******e 发帖数: 133 | 15 职位是Marketing Analyst
Base 120K差点
Bonus 15%
RSU 105K over 4 years
No sign in bonus
我:不是engineer, 统计的master, 有7年工作经验。 这个Base和其他中小型公司同类
职位比也不算高,一个我不是engineer, 另外工资是比较我目前公司的pay, 当时想先
搬来湾区所以take了目前公司的low pay。
发个面经给去面试的人参考一下时间流程. 从recuiter联系我到offer一共六周时间:
07/21 Apple recruiter发邮件问我对一个职位有没有兴趣
07/22 Apple recruiter phone screen
07/28 phone interview with hiring manager
08/04 On site interview, 一共三小时,包括recruiter面6个人,每人半小时
08/05 hiring manager回复我的thank you letter, 说所有的人给我的评价都很好,
recruiter会联系我
08/11 recui... 阅读全帖 |
|
m********t 发帖数: 94 | 16 自己顶一下
这种东西怎么做拟合或者预测啊?
用户 |
|
m********t 发帖数: 94 | 17 自己再定下
还是大家都觉得一个linear regression就足够了?
自己顶一下
这种东西怎么做拟合或者预测啊?
用户 |
|
E********t 发帖数: 418 | 18 请问如果是一个 ARIMA(4,0,2) model,其中 AR(2) AR(3)term not significant,可以
不放在model里面吗?我的意思是只用 AR(1) AR(4) MA(1) MA(2)term 去 forecast
还有请问 proc armria 怎么才能显示 RMSEF (roost mean squared error) 呢?谢
谢! |
|
g******2 发帖数: 234 | 19 not an expert, just want to throw bricks to lure jades (pao zhuan yin yu):
1. calculate linear correlation?
2. build a linear model x ~ y using only part of the data, and use other
part of the data as a validation data set.
3. a) fit a time series model for time series x (arima etc.), record the
fitted values
b) randomly swap n elements of time series x with y, refit the time
series model, record fitted values, calculate the distance (MSE etc.)
between the fitted values from b) and those from... 阅读全帖 |
|
s***h 发帖数: 26 | 20 Time Series很渣,Proc Arima不开交互模式不能用。并且ods有些也用不了我记不太清
楚了
建议你用SAS University的升级版,odamid.oda.sas.com
你说的两个model都没问题 |
|
w*******e 发帖数: 182 | 21 有这样一组数据,一个网站每天早上,上午,下午和晚上四个时间段的流量,这个流量
有很强的seasonality pattern:weekday(比如:周一和上周一,下周一)之间,同一
时间段之间。这种怎么simulate呢? 比如用ARIMA model的话,怎么确定 p,n,q的值呢
? |
|
a*****i 发帖数: 1045 | 22 auto.arima, R 里面的,可以自动选。不过通常效果不是,老师上课的时候说不建议。 |
|
S*********1 发帖数: 105 | 23 first difference all non-stationary variables, then build a regression with
ARIMA errors |
|
A****1 发帖数: 33 | 24 what does its time series look like?
does it have periodic waves, trends, or intervals? Couple of things need to
be considered. |
|
l******n 发帖数: 9344 | 25 来自主题: DataSciences版 - 也问个模型 Xnew直接arima就好了 |
|
|
|
c****t 发帖数: 19049 | 28 不同月的新用户们之间有什么关系?如果没什么关系这不是一个严格的time-series问题
用户 |
|
m********t 发帖数: 94 | 29 当然是相对稳定 但是肯定有一定波动的
我其实想知道有没有一个很直接的方法把这两列数字扔进去 就能出来结果的model
不同月的新用户们之间有什么关系?如果没什么关系这不是一个严格的time-series问题
用户 |
|
a***g 发帖数: 2761 | 30 来自主题: DataSciences版 - 请教一个题 你这个例子 典型arima啊 |
|
T*****u 发帖数: 7103 | 31 来自主题: DataSciences版 - 请教一个题 arima,这种周期性的预测都已经成熟了吧。 |
|
w**2 发帖数: 147 | 32 我最近的一个project是modelling inflation rate。好惨,那个time series里面有很
多regime change产生的影响,一看就不是个正常的time series。
我最后fit了一个arima model 加 arch/garch 效果还行吧。你的time series如何
irregular,能说说看嘛 |
|
w**2 发帖数: 147 | 33 time series, 先检查stationarity,seasonality,然后试着fit一个arma, arima or
sarima之类的。 |
|
t******g 发帖数: 2253 | 34 我来问几个基本的问题
1) ARMA的p, q 如何选择
2) 如果有seasonality怎么处理
3) 什么是ARIMA
4) 什么是exponential smoothing model |
|
|
b*****s 发帖数: 11267 | 36 6000/500=12吧? 10.2是怎么算出来的?
要波动小就是取平均值呀,最简单的model,assume independent
如果assume dependent,那就是arima,mcmc之类的吧,但是这个一般只预测 near
future,而且是按天来算的。预测一个月,差别我觉得未必比第一个号
coming |
|