Okay, let's first consider the distribution of matter and how it changes. Is star formation "creating normal matter"? No, it just recycles the matter that was already there from the interstellar medium. (In fact, because star death does not return 100% of that material back, we would say the ISM is being depleted over time, as more mass is converted from gas to stars to stellar remnants, at least for as long as star formation continues). But the point is, the total amount of normal matter is the same.
Another problem is that because there are more stars per unit volume of space near the galactic center than the outer regions, the hypothesis that star formation helps explain dark matter by creating new matter does not work. What we would need is more matter being created at larger distances, to even things out to produce a flat rotation curve. Furthermore, we still need to explain what that matter is and why we don't see it. We know that regular matter is not distributed in such a way as to be consistent with the rotation curves.How do we know that?
Consider what we are able to measure. Stars are easy. They are luminous and we know how to measure their masses, and thus we can get a pretty good measure of a galaxy's stellar mass. Gas is also luminous, but at different wavelengths. We can survey ionized gas, which occurs in star formation regions, by their emission lines. We can also observe neutral molecular gas by 21cm wavelength emission. And we can observe hot gas (like the galactic corona) in X-rays. In short, we have methods of surveying all the types of regular matter, and we know pretty well how much of it there is and how it is distributed.
When we do this, and plot how this matter is distributed as a function of distance, and then apply the laws of gravity (basically Kepler's Third Law) to predict how quickly things should orbit as a function of distance, we get a specific predicted curve, and that curve drops off with distance, much like how the orbital velocity of planets in a solar system drops off with distance.
But what we actually observe that the velocity vs. radius curve doesn't drop off as expected. It is essentially flat out to large radii -- even out to the very edge of the galactic disk. What this means is that either our understanding of gravity is wrong, or that there really is more matter there. In fact, we can also extend this beyond the galactic disk -- by measuring the velocities (velocity dispersion) of the gas in the galactic corona. We get the same result -- there must be more matter even out there than what we are able to detect by electromagnetic observations.
There are only two options -- either our models of gravitation must be adjusted at large distances (or slow accelerations)... or there really is more matter there than what we can see: "dark matter". Both of these are ad hoc hypotheses, but they make further predictions which we can test. It turns out that the dark matter hypothesis has much greater predictive success than the modified gravitation.
(There is a third option which is that there is more regular matter but in the form of "dark" components that we already know about -- like rogue planets and black holes, but when we go through the maths and the constraints from observations of how many of these things there can be out there, there are not enough. Dark matter as an exotic new type of particle is still necessary.)
Dark matter naturally predicts that there should be places in the universe where it gets separated from normal matter (such as during galactic collisions, like the Bullet Cluster), and indeed this is what we observe. Another check is that because we measure the spatial curvature of the universe to be flat (mass density ~= critical density), yet what we observe for the density of regular matter is only about 5% of that, then adding together the amount of dark matter and dark energy needed to explain the formation of structure in models of cosmological simulations ought to bring us to 100% of the critical density. And indeed it does. This isn't number fudging -- it is different observational tests yielding consistent results when they easily could have worked out differently.
Astronomers use the LCDM (Dark Energy and Cold Dark Matter) model with Ωm
~0.3 and ΩΛ
~0.7, meaning the universe is 30% matter, of which only ~5% is normal matter. 25% is dark matter, and the remaining 70% is dark energy. These numbers make specific predictions for a variety of cosmological observations, and they all happen to agree very well.
That is why this model is considered the concordance -- or "most accepted and used" model of cosmology.