Have you analyzed the correlation between the other dependent variables? Some times this type of results can be attributed to the presence of multicollinearity in your model (a high correlation between independent variables).

Alhanouf - 

Yes, collinearity can cause this.  Perhaps you could try some alternative models, and compare their performances.  R-square is not what you want for this.  You could try using variables that the subject matter suggests, and try to avoid those that give you basically the same information.  (Principle components might help, but then interpretation is more vague.)  Graphical residual analysis can be useful, and holding out some test data can help avoid overfitting.

Something in here might be helpful:

 https://www.researchgate.net/publication/313925533_Comparing_Model_Performances_Graphically

You might also consider the following, by Galit Shmueli:

https://www.researchgate.net/publication/48178170_To_Explain_or_to_Predict

The example in the appendix is helpful. 

The estimated variance of the prediction error is useful, for individual cases and for totals for finite populations.  Because sigma has to be estimated, any bias in a model will influence the estimate of sigma, and this will impact the estimated variance of the prediction error.  Thus it can become a good overall indicator of accuracy. 

Cheers - Jim

Yes, this is very likely because of multicollinearity. Try to estimate VIF (variance inflation factor) for each variable in your data. There are a lot of ways in R or Python to do that.

https://www.displayr.com/variance-inflation-factors-vifs/

Shortly - big VIF indicates that this variable heavily correlates to some other variable.

Then simply drop the variable which has the biggest VIF and try to build the model without it. Maybe the model would be more plausible. If you still don't like it, you can repeat the procedure.

Good luck!

Thanks everyone i estimited vif and it resulted high vif in 3 variables 😕