What is the difference between similarity measurement in time series and similarity search in time series ? I am abit confused with these two terms.
to my understanding. Similarity search is the process of obtaining similar time series using similarity measure such as euclidean distance, DTW, EDR, EDP and etc.
Then what is similarity measurement ?
I am doing a project titled similarity measurement between time series. I am totally confused regarding my aim and objectives of my project.
After weeks of reading journals and papers, I come up with a plan for my project. Please clarify whether I am doing the right thing.
I plan to use 2 popular similarity measure which are euclidean distance and Dynamic time warping as the similarity measure. There will be a time series database. When a query is made towards the time series database, The distance between query data and each of the data in the database will be computed and compared. If the distance is within a threshold (How do I determine this threshold ? ) , this time series will be counted as one of the candidates that has similar pattern with the query data. Different type of time series data will be used such as stock data, leaf data, product sale data.
Up to this point, my question is how do i do performance evaluation on these two similarity measures ? What is the best evaluation to compute it's accuracy and efficiency?
Then, I shall apply dimensionality reduction method such as DFT or DWT on the data. My question here is do I have to do indexing on the reduced data set ? I will then compare two of the similarity measure performance.
Given my plan above, I wish to do matlab to do the implementation. However, I realised that if i wanna do indexing, matlab may not be too suitable. What solution do I have on this matter ?