What are some good methods/ algos/ models for time series forecasting (apart from the moving average methods) for sales prediction in an apparel selling store.?
Exponential smoothing model are great for doing univariate demand forecasting. If you are doing forecasting in R then you can look at Forecast package and if you are doing this in SAS then you can look at proc esm.
Here are a few models worth trying. @Abhishek has already mentioned about exponential smoothing. You can also try linear and quadratic smoothing or ARIMA parametric time series modeling.
If you have more details about the problem or outcome from something you have already tried, we might be able to give a more specific answer.
Basically I have data set with time stamps and sales (no. of quantities sold) and the trend is such that there’s a spike on specific days, i.e, weekends and festive seasons. I have to make sure that I capture this trend in my predictions too. Also there is a slight increase in the sales for the overall period, i.e, the slope of the linear trendline is positive. Don’t have any other information as such so can’t use machine learning algos for prediction (or can I?).
I am willing to use R and Python for the same.
In general, there are several methods to be used for forecasting time series. like ARIMA,Neural Network,Exponential Smoothing State Space Model,Dynamic Regrssion Models…Exponential Smoothing state space modeling framework is introduced for forecasting complex seasonal time series such as those with multiple seasonal periods, high frequency seasonality, non-integer seasonality.
@ayush_sharan There are many techniques for forecasting like @karthiv has mentioned. I suggest you to try simple methods where you can apply your understanding of data. Time series data consists of mainly four components - trend, season, cycle and noise (Refer). I would recommend you to estimate the trend and seasonal component (Ex:- day of week ) from your data and use these to forecast future. To estimate trend you can use regression (try linear) and for estimating seasonal component use a method called average seasonal index extractorSIE. All these you can do in spreadsheets.
Hope that helps.
Exponential Smoothing State Space Model.
A key feature of the proposed trigonometric framework is its ability to model both linear and non-linear time series with single seasonality, multiple seasonality, high period seasonality, non-integer seasonality and dual calendar effects and trend.In addition, the framework consists of a new estimation procedure which is sufficiently general to be applied to any innovations state space model. By relying on maximum likelihood estimation, it avoids the ad hoc start up choices with unknown statistical properties commonly used with exponential smoothing.
you can use TBATS Function in R
Thanks all for the help. Was not able to reply earlier but your suggestions have been very helpful.
Need to ask 3 questions:
Q-1) I want to forecast automotive industry growth? What should be the algorithm to be used?
Q-2) Also for forecasting company internal sales data what all factors and time series technique can be used?
Q-3) Which technique is used when we have explaining variables(more than 1) along with the sales data?