Base Conversion With Strings

Python 2, 115 114 106 105 94 bytes

Golfing suggestions welcome. Try it online!

Edit: -9 bytes thanks to mbomb007. -2 bytes thanks to FlipTack.

def a(n,f,t,d,z=0,s=''):
 for i in n:z=z*f+d.find(i)
 while z:s=d[z%t]+s;z/=t
 print s or d[0]

Ungolfed:

def arbitrary_base_conversion(num, b_from, b_to, digs, z=0, s=''):
    for i in num:
        z = z * b_from + digs.index(i)
    while z:
        s = digs[z % b_to] + s
        z = z / t
    if s:
        return s
    else:
        return d[0]

Seriously, 50 bytes

0╗,╝,2┐,3┐,4┐╛`4└í╜2└*+╗`MX╜ε╗W;3└@%4└E╜@+╗3└@\WX╜

Hex Dump:

30bb2cbc2c32bf2c33bf2c34bfbe6034c0a1bd32c02a2bbb60
4d58bdeebb573b33c0402534c045bd402bbb33c0405c5758bd

I'm proud of this one despite its length. Why? Because it worked perfectly on the second try. I wrote it and debugged it in literally 10 minutes. Usually debugging a Seriously program is an hour's labor.

Explanation:

0╗                                                  Put a zero in reg0 (build number here)
  ,╝,2┐,3┐,4┐                                       Put evaluated inputs in next four regs
             ╛                                      Load string from reg1
              `         `M                          Map over its chars
               4└                                   Load string of digits
                 í                                  Get index of char in it.
                  ╜                                 Load number-so-far from reg0
                   2└*                              Multiply by from-base
                      +                             Add current digit.
                       ╗                            Save back in reg0
                          X                         Discard emptied string/list.
                           ╜                        Load completed num from reg0
                            ε╗                      Put empty string in reg0
                              W                W    While number is positive
                               ;                    Duplicate
                                3└@%                Mod by to-base.
                                    4└E             Look up corresponding char in digits
                                       ╜@+          Prepend to string-so-far.
                                                      (Forgetting this @ was my one bug.)
                                          ╗         Put it back in reg0
                                           3└@\     integer divide by to-base.
                                                X   Discard leftover 0
                                                 ╜  Load completed string from reg0
                                                    Implicit output.

CJam, 34 bytes

0ll:Af#lif{@*+}~li:X;{XmdA=\}h;]W%

Input format is input_N alphabet input_B output_B each on a separate line.

Run all test cases.

Explanation

0     e# Push a zero which we'll use as a running total to build up the input number.
l     e# Read the input number.
l:A   e# Read the alphabet and store it in A.
f#    e# For each character in the input number turn it into its position in the alphabet,
      e# replacing characters with the corresponding numerical digit value.
li    e# Read input and convert to integer.
f{    e# For each digit (leaving the base on the stack)...
  @*  e#   Pull up the running total and multiply it by the base.
  +   e#   Add the current digit.
}
~     e# The result will be wrapped in an array. Unwrap it.
li:X; e# Read the output base, store it in X and discard it.
{     e# While the running total is not zero yet...
  Xmd e#   Take the running total divmod X. The modulo gives the next digit, and
      e#   the division result represents the remaining digits.
  A=  e#   Pick the corresponding character from the alphabet.
  \   e#   Swap the digit with the remaining value.
}h
;     e# We'll end up with a final zero on the stack which we don't want. Discard it.
]W%   e# Wrap everything in an array and reverse it, because we've generated the 
      e# digits from least to most significant.

This works for the same byte count:

L0ll:Af#lif{@*+}~li:X;{XmdA=@+\}h;

The only difference is that we're building up a string instead of collecting everything on the stack and reversing it.