IoT Analytics - Time series


#1

Hi All,

We will receive data of Car parking, lighting and environment for every minute from sensors. We want to fit the best curve for the real-time data. What are the best algorithms for smooth curve fitting? Based on the historical data with the time, we should able to predict the future occupancy rate and an environment.

Can any one suggest me any smart models for forecasting ??


#2

Hi @innamurisrikanth,

I would suggest that before jumping on to Machine Learning algorithms for curve fitting, first and foremost, dig into the data (visualization, slicing, dicing etc) and try to understand it. See if you can find the hidden patterns. Then move on to using more complex procedures.

PS: Go through this and this resource for time series modelling.


#3

Hi @innamurisrikanth

There are a number of curve fitting algorithms available

Let me take help of the smooth.spline function in R

set.seed(100)
x <- runif(100)
y <- runif(100)
plot(y,x)
model <- smooth.spline(y,x)
lines(model,col="blue")

set.seed(100)
x <- seq(1,10)
y <- x^3 - x^2 + 5
plot(y,x)
model <- smooth.spline(y,x)
lines(model,col="blue")

Hope this helps

Regards,
Anant