ML experts

[erragulabi]

kind of new to this. I am running some examples and notuced most of them are used to compare data. Like we have year 2019 data and year 2020 data. We create a trained model using 2019 and predict 2020 data and comparing. What if we want use both 2019 and 2020 year data and predtict future year data? Can someone help in pointing to right direction?

[erragulabi]

LTT 

[Dippam]

51 minutes ago, erragulabi said:

kind of new to this. I am running some examples and notuced most of them are used to compare data. Like we have year 2019 data and year 2020 data. We create a trained model using 2019 and predict 2020 data and comparing. What if we want use both 2019 and 2020 year data and predtict future year data? Can someone help in pointing to right direction?

I am not an expert in ML.

At a high level,

Train data: 2019 & 2020 (used for training the model)

Test data: Future data. ( if u have it available already to test the model performance)

Look into concepts like class imbalance, weighted classes, time series data

[erragulabi]

14 minutes ago, Dippam said:

I am not an expert in ML.

At a high level,

Train data: 2019 & 2020 (used for training the model)

Test data: Future data. ( if u have it available already to test the model performance)

Look into concepts like class imbalance, weighted classes, time series data

Thanks bro..may be my assumption is wrong here. How we will have future data in our hands? Might have to hit the basics. But imagine if we have 100 yr weather data in our hand and we know max rainfall happened in a area and with that data in hand, cant we predict next 10 year weather? 

[k3I]

3 hours ago, erragulabi said:

kind of new to this. I am running some examples and notuced most of them are used to compare data. Like we have year 2019 data and year 2020 data. We create a trained model using 2019 and predict 2020 data and comparing. What if we want use both 2019 and 2020 year data and predtict future year data? Can someone help in pointing to right direction?

 

what kind of data is this? If the data follows the same distribution over the years, you can just combine it and train it to predict new data.

[k3I]

.

[erragulabi]

lets assume sales data..and model using linear regression

[k3I]

4 minutes ago, erragulabi said:

lets assume sales data..and model using linear regression

need a little more context here( what exactly are you trying to predict? is it a vector?). anyway, if you have ground truth for both the years and if they follow the same distribution, just extend the dataset for both the years and train on it.

[jambalhaatraja]

3 hours ago, erragulabi said:

kind of new to this. I am running some examples and notuced most of them are used to compare data. Like we have year 2019 data and year 2020 data. We create a trained model using 2019 and predict 2020 data and comparing. What if we want use both 2019 and 2020 year data and predtict future year data? Can someone help in pointing to right direction?

forget future data man.. focus on historical and event based predictions...much of the industry is still garnered around traditional data...even tesla cant predict how will a driver behave based on his prior driving experience

ML is just building a model on old data..extracting an equation with weights and intercepts...use that to predict incoming data..whether it is for regression clustering recommendation computer vision NLP reinforcement learning etc