As shown in the above, this is a refrigerator in the real-time temperature curve, each 1 minute records a data, you can see the temperature data of ups and downs regularly



But as shown above, the normal use of the refrigerator, the door opened and there will be a temperature rise, now want to achieve in the daily data analysis, by the system automatically determine the data like this belongs to the normal temperature drop caused by everyday use, is not due to power outages or abnormal events, caused by damage of refrigerator technology on how to carry out the analysis of these data points?

CodePudding user response:

From temperature is look not to come out, need auxiliary information, such as in the timing of the alarm has a history of doors

CodePudding user response:

Discrete time sequence data characteristic, this thing although I would also point, but I belong to half a bottle of water, don't give advice,

You should go to math or automation system signal processing to modeling (generally is timing window, difference equation and statistical characteristic model), artificial intelligence, also line (this need not modeling, you give large quantities of data to make them automatic binary statistical classification can also)

CodePudding user response:

If the requirements of real-time, can put the data in kafka or special temporal database, behind that can match the spark, flink do external distributed computing, after all, even if there is a special talent for you do the math model, a two sets of data, easy one thousand units ten thousand units is bad

CodePudding user response:

Reference

if the requirements of real-time, can put the data in kafka or special temporal database, it can match the spark behind, flink do external distributed computing, after all, even if there is a special talent for you do the math model, a two sets of data, easy one thousand units ten thousand units is bad forget it

Thanks for a great god, and have a look at these professional term is dizzy

CodePudding user response:

reference 4 floor sunny2u07 response:

reference

if the requirements of real-time, can put the data in kafka or special temporal database, behind that can match the spark, flink do external distributed computing, after all, even if there is a special talent for you do the math model, a two sets of data, easy one thousand units ten thousand units is bad forget it

Thanks for a great god, and have a look at these professional term is dizzy

So, you can apply for a IOT to BAT first test accounts, experience the BAT have such things as how to play

For example: baidu swan-shaped system
https://cloud.baidu.com/solution/iot/index.html

CodePudding user response:

reference 5 floor wanghui0380 reply:

Quote: refer to 4th floor sunny2u07 response:

Reference

if the requirements of real-time, can put the data in kafka or special temporal database, it can match the spark behind, flink do external distributed computing, after all, even if there is a special talent for you do the math model, a two sets of data, easy one thousand units ten thousand units is bad forget it

Thanks for a great god, and have a look at these professional term is dizzy

So, you can apply for a IOT to BAT first test accounts, experience the BAT have such things as how to play

For example: baidu swan-shaped system
https://cloud.baidu.com/solution/iot/index.html

thank you very much